OpenAI pricing, complete breakdown
Verified 2026-05-27, cross-checked against OpenAI pricing page, litellm, openrouter
OpenAI's current model lineup is led by GPT-5.5 at $5.00 per million input tokens and GPT-5.4 at $2.50. For high-efficiency applications, GPT-5 Nano offers an entry rate of $0.05 per million input tokens, while specialized reasoning models like o3-deep-research are priced at $5.00 for complex tasks. Developers requiring maximum performance can access GPT-5.5 Pro at $30.00 per million input tokens. This page helps you navigate these metered API rates alongside the expanding suite of ChatGPT subscription tiers to optimize your total AI spend.
GPT-5 Nano is the most affordable entry point at $0.05 per 1M input tokens.
How OpenAI's pricing universe works
OpenAI operates a multi-track pricing strategy to balance high-margin developer growth with predictable consumer revenue. Frontier model companies require massive capital for compute, so they offer API access for builders who need granular control and subscriptions for end-users who need a ready-made interface. This allows OpenAI to capture value from individual power users, collaborative teams, and large-scale programmatic integrations simultaneously. By diversifying access modes, they ensure that the same underlying models can serve a $20/month hobbyist and a $50,000/month enterprise application.
API (per-token, metered)
For: Developers, technical teams, startups building products on top of OpenAI
- Pay only for tokens consumed
- Full model lineup including batch, caching, long context
- Programmatic via SDKs
When to use: When integrating OpenAI into your own product or running variable batch workloads
Best for: Builders with metered or unpredictable usage
Consumer subscriptions (Plus, Pro tiers)
For: Individuals using OpenAI directly for writing, coding, research, analysis
- Fixed monthly fee
- Generous usage caps
- Web/desktop/mobile apps
- Often includes newer models first
When to use: When using OpenAI as a daily-driver AI assistant rather than building on it
Best for: Solo professionals, knowledge workers, vibe coders
Business/Team plans
For: Teams of 2-200 needing shared workspaces, admin controls, SSO
- Per-seat billing
- Centralized billing
- Admin & audit controls
- Shared projects and custom workspace GPTs
When to use: When deploying OpenAI across a team that does NOT need API integration
Best for: Mid-size organizations adopting AI for internal productivity
Enterprise (custom contract)
For: Large organizations with procurement requirements, compliance needs, or volume-discount leverage
- Custom pricing and limits
- SLAs
- DPAs and BAAs
- Dedicated support
- Data residency in ten regions
When to use: When per-seat or per-token pricing exceeds ~$50K/year, or when compliance/contractual needs require it
Best for: Enterprises with procurement-led adoption
Cloud marketplaces (Azure OpenAI)
For: Organizations with existing cloud commits or strict data-residency requirements
- Same models, slightly different pricing (often parity or small premium)
- Counts toward existing cloud spend commits
- Stays within cloud's data-protection boundary
When to use: When you already burn down EDP/MACC/CCC commits and prefer single-bill
Best for: Cloud-committed enterprises
Which one should you pick? If you are building a software product, use the API for metered control. For personal productivity, choose ChatGPT Plus or the Pro $100/$200 tiers depending on your research needs. Teams should adopt the ChatGPT Team or Business ChatGPT & Codex plans for shared workspaces. Large-scale organizations with strict compliance requirements should opt for ChatGPT Enterprise or the Azure OpenAI Service.
⭐ Most popular OpenAI products by user type
Skip the deep dive. These are the natural fits for each user type.
⭐
Casual User
ChatGPT Free
$0
Natural fit for users who need basic reasoning and occasional access to GPT-4o mini without a monthly commitment.
+ Light Paid
See full profile ↓
⭐
Power User
ChatGPT Plus
$20/mo
Designed for individuals who require higher message caps on GPT-4o and access to DALL-E, Data Analysis, and Custom GPTs.
+ Heavy User
See full profile ↓
⭐
Advanced Researcher
ChatGPT Pro
$200/mo
Optimized for professionals who need unlimited access to the o1-pro mode for complex scientific or mathematical reasoning.
See full profile ↓
⭐
Small Business
ChatGPT Team
$25-30/user/mo
A logical choice for groups of 2+ who need a shared workspace, admin controls, and data exclusion from training by default.
+ Technical Teams
See full profile ↓
⭐
Software Developer
OpenAI API
Pay-as-you-go
The standard for builders integrating models directly into applications with granular control over token usage and model versions.
See full profile ↓
🎁 Current promos and time-sensitive deals
What's active right now. Auto-hides expired items.
OpenAI for Startups / VC Partner Program
Up to $100,000 in free API credits for GPT-5.2 and other models, usage tier upgrades, and access to new agent infrastructure.
YC Startup Batch Credits
$2 million in API credits for startups accepted into Y Combinator (e.g., S25 or S26 batches) for model training and inference.
Azure OpenAI Provisioned Throughput Units (PTU) Discount
Reserved capacity pricing reportedly 18% to 34% below pay-as-you-go rates for flagship models.
Promotional credits are typically subject to OpenAI's standard API terms and may have specific expiration windows (often 6-12 months) once issued.
📅 What changed in the last 30 days
Populated from aicost_price_changelog. Hides automatically when no recent events.
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-realtime-2 · costPerMinute changed
Now 0.034000
Now 0.034000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
openai-gpt-realtime-whisper added to catalog
New model added: GPT-Realtime Whisper — inserted by aicost-merge-new-models
New model added: GPT-Realtime Whisper — inserted by aicost-merge-new-models
·
openai-gpt-image-2 added to catalog
New model added: GPT-Image 2 — inserted by aicost-merge-new-models
New model added: GPT-Image 2 — inserted by aicost-merge-new-models
·
openai-gpt-realtime-2 added to catalog
New model added: GPT-Realtime 2 — inserted by aicost-merge-new-models
New model added: GPT-Realtime 2 — inserted by aicost-merge-new-models
·
openai-gpt-realtime-translate added to catalog
New model added: GPT-Realtime Translate — inserted by aicost-merge-new-models
New model added: GPT-Realtime Translate — inserted by aicost-merge-new-models
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-4-pro · longContextThreshold changed
270000.000000 → 200000.000000
270000.000000 → 200000.000000
·
gpt-5-4 · longContextThreshold changed
270000.000000 → 200000.000000
270000.000000 → 200000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-4-mini · longContextThreshold changed
Now 270000.000000
Now 270000.000000
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5-pro · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-5 · longContextThreshold changed
272000.000000 → 270000.000000
272000.000000 → 270000.000000
·
gpt-5-4 · longContextThreshold changed
270000.000000 → 200000.000000
270000.000000 → 200000.000000
·
gpt-5-4-pro · longContextThreshold changed
270000.000000 → 200000.000000
270000.000000 → 200000.000000
·
gpt-5-4-mini · longContextThreshold changed
Now 270000.000000
Now 270000.000000
Every OpenAI product, profiled
For each product, what it's for, who picks it, what to watch out for, pros and cons, and what we tell our consulting clients.
consumer free
ChatGPT Free
Free
Free forever
Target users
casual users, students, curious explorers
Typical uses
- Basic writing assistance and brainstorming
- General knowledge queries
- Testing GPT-5-mini capabilities
- Casual image generation with DALL-E
Why pick it
The best entry point for individuals who need occasional AI assistance without a financial commitment.
Key features
- Access to GPT-5-mini and GPT-5-nano
- Limited access to flagship GPT-5 model
- Basic data analysis and file uploads
- Custom GPTs (limited usage)
- Web browsing and vision capabilities
⚠ Marketing gimmicks to watch
Dynamic Usage Caps
OpenAI does not publish a fixed message limit for free users; caps vary based on real-time server demand.
Impact: Expect to be downgraded to lower-tier models (like GPT-5-nano) during peak hours without warning.
Model Gating
Flagship models are available but heavily restricted, often reverting to 'mini' models after a few messages.
Impact: Not suitable for complex, multi-step projects that require consistent high-reasoning performance.
Pros
- Zero cost for life
- Access to state-of-the-art 'mini' models
- Includes mobile app and voice mode
Cons
- Frequent 'at capacity' messages during peak times
- No priority access to new features
- Data is used for training by default
Insider view
ChatGPT Free is essentially a 'loss leader' designed to showcase OpenAI's ecosystem. It is perfectly adequate for basic tasks, but the inconsistent performance during high-traffic periods makes it frustrating for professional workflows.
Max bang for buck
Use the mobile app for voice-to-text brainstorming, which is often less restricted than the web-based reasoning interface.
🔒 Training-on-your-data policy
OpenAI may use your content to improve models. Users can opt-out via 'Data Controls' in settings. Source: https://help.openai.com/en/articles/7730893-data-controls-faq
🔄 Migration path
Upgrade when:
You hit message limits more than twice a day or need Advanced Voice Mode.
Downgrade when:
N/A (lowest tier)
Switch vendor when:
You require a larger context window for free (consider Google Gemini).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Standard individual usage | varies | varies | No hidden fees for standard usage. |
consumer entry paid
ChatGPT Go
$8/mo · $96/yr
Monthly billing only
Target users
light professionals, mobile first users, budget conscious creatives
Typical uses
- Daily productivity assistance
- Enhanced mobile AI interactions
- Moderate GPT-5 usage without Plus cost
Why pick it
Designed for users who have outgrown the Free tier but don't need the heavy-duty limits of Plus.
Key features
- Higher message caps than Free tier
- Priority access during peak times
- Standard GPT-5 access
- Full access to Custom GPTs
- Advanced Voice Mode (limited)
⚠ Marketing gimmicks to watch
The 'Middle Child' Gap
This tier often lacks the 'Advanced' features like DALL-E 3 or high-res vision found in Plus.
Impact: Check feature lists carefully; you may still hit walls on creative tasks.
No Annual Discount
Unlike Team or Business tiers, Go is strictly month-to-month.
Impact: Costs $96/year with no way to lower the price via commitment.
Pros
- Affordable entry into paid AI
- Better reliability than Free tier
- Access to the GPT Store
Cons
- Still has meaningful usage caps
- Lacks the 'Pro' tools of the $20 tier
- No annual billing option
Insider view
ChatGPT Go is OpenAI's response to users who found $20/month too steep. It’s a great 'Goldilocks' tier for students or light office workers who just want the AI to work when they need it.
Max bang for buck
If you primarily use AI for text-based productivity, Go provides 80% of the value of Plus for 40% of the price.
🔒 Training-on-your-data policy
Content is used for training by default unless opted out in settings. Source: https://openai.com/policies/privacy-policy
🔄 Migration path
Upgrade when:
You need DALL-E 3 for image generation or hit message caps on GPT-5.
Downgrade when:
You find yourself using the tool less than once every few days.
Switch vendor when:
You need better integration with Google Workspace (switch to Gemini).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Individual light pro usage | $8 | $96 | Billed monthly at $8. |
consumer flagship
ChatGPT Plus
$20/mo · $240/yr
Standard consumer rate
Target users
power users, developers, content creators
Typical uses
- Complex coding and debugging
- High-quality image generation with DALL-E 3
- Advanced data analysis on large datasets
- Frequent use of Advanced Voice Mode
Why pick it
The industry standard for individuals who want the full power of OpenAI's latest frontier models with high usage limits.
Key features
- Early access to new features (e.g., SearchGPT, Sora)
- 5x more messages on GPT-5 compared to Free
- DALL-E 3 image generation
- Advanced Voice Mode with low latency
- Full access to GPT Store and Custom GPTs
⚠ Marketing gimmicks to watch
Soft Limits
OpenAI advertises 'high limits' but reserves the right to throttle users during periods of extreme demand.
Impact: You may see a message saying you've reached your limit for the next few hours, even as a paying subscriber.
Feature Gating
New features like 'o1-preview' or 'o3-deep-research' often have much lower caps than standard GPT-5.
Impact: Don't assume 'Plus' means 'Unlimited' for the newest, most expensive models.
Pros
- First-in-line for frontier model updates
- Excellent multimodal capabilities (vision/voice/image)
- Large ecosystem of Custom GPTs
Cons
- No annual discount for individuals
- Usage caps still exist on reasoning models
- Data training is opt-out, not off-by-default
Insider view
Plus remains the benchmark for consumer AI. While competitors like Claude Pro offer better writing, Plus offers a more versatile 'Swiss Army Knife' experience with its vision and voice tools.
Max bang for buck
Use the 'Custom Instructions' feature to reduce repetitive prompting, which saves time and mental overhead.
🔒 Training-on-your-data policy
OpenAI uses content to improve models unless the user opts out in settings. Source: https://help.openai.com/en/articles/7730893-data-controls-faq
🔄 Migration path
Upgrade when:
You need to share a workspace with a team or require data privacy by default.
Downgrade when:
You only use the AI for basic text tasks that GPT-5-mini can handle.
Switch vendor when:
You need a 200k+ context window for large document analysis (switch to Claude).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Standard power user | $20 | $240 | Billed monthly at $20. |
consumer power low
ChatGPT Pro $100
$100/mo · $1200/yr
High-tier individual plan
Target users
professional researchers, heavy reasoning users, AI first developers
Typical uses
- Deep research using o3-deep-research models
- Large-scale code architecture planning
- Complex logical reasoning tasks
Why pick it
Designed for users whose workflows rely on OpenAI's most computationally expensive reasoning models.
Key features
- Significantly higher caps on o-series models
- Priority access to 'Pro' versions of flagship models
- Extended context handling for reasoning tasks
- All features of ChatGPT Plus included
⚠ Marketing gimmicks to watch
Reasoning Token Inflation
Reasoning models generate internal tokens that count against your limits but aren't visible.
Impact: Even with higher caps, complex queries can 'burn' through your daily allowance faster than expected.
The 'Pro' Label Confusion
OpenAI uses 'Pro' for both this $100 tier and the $200 tier, differing only in usage volume.
Impact: Ensure you are selecting the correct dollar-amount tier for your specific volume needs.
Pros
- Massive increase in reasoning model capacity
- Reduced latency on frontier models
- Ideal for technical power users
Cons
- Very high price point for an individual
- No additional 'features' over Plus, just higher limits
- No annual discount
Insider view
This is a niche tier for people who literally 'live' in ChatGPT. If you aren't hitting the $20 Plus limits daily, this is a waste of money. If you are, the productivity gain from not being throttled is worth the $100.
Max bang for buck
Only subscribe if your work involves 'o-series' models for more than 4 hours a day.
🔒 Training-on-your-data policy
Refer to vendor privacy policy; usually follows Plus (opt-out). Source: https://openai.com/policies/privacy-policy
🔄 Migration path
Upgrade when:
You are consistently hitting 'limit reached' on o1 or o3 models.
Downgrade when:
You find yourself using standard GPT-5 more than the reasoning models.
Switch vendor when:
You need specialized coding tools like GitHub Copilot or Cursor.
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Heavy reasoning user | $100 | $1,200 | Billed monthly. |
consumer power high
ChatGPT Pro $200
$200/mo · $2400/yr
Maximum individual tier
Target users
AI power users, independent consultants, quantitative analysts
Typical uses
- Unrestricted reasoning model usage
- High-volume data synthesis
- Full-day AI-assisted development
Why pick it
The highest possible usage limits available for an individual account without moving to a Team or Enterprise plan.
Key features
- Maximum caps on o-series and GPT-5-pro models
- Highest priority in the global compute queue
- Access to all experimental tools and models
- Includes all Plus features
⚠ Marketing gimmicks to watch
The 'Unlimited' Illusion
Even at $200, usage is not truly 'unlimited'; it is 'uncapped for normal professional use'.
Impact: Automated scraping or bot-like behavior will still trigger account reviews.
Individual vs Team Gating
This tier is for one person; you cannot legally share it with a partner or assistant.
Impact: If you need two people, ChatGPT Team ($60/mo total) is significantly cheaper.
Pros
- Virtually eliminates 'limit reached' anxiety
- Best-in-class performance for reasoning tasks
- No need to manage API credits
Cons
- Extremely expensive for a single subscription
- Diminishing returns for most users
- No team management features
Insider view
This tier is effectively 'unlimited' for any human user. It's designed for the top 0.1% of users who use AI as their primary work interface. For everyone else, the $20 or $100 tiers are better value.
Max bang for buck
If you are spending more than $200/mo on API credits for personal research, this plan is a massive cost-saver.
🔒 Training-on-your-data policy
Refer to vendor privacy policy; usually follows Plus (opt-out). Source: https://openai.com/policies/privacy-policy
🔄 Migration path
Upgrade when:
You are a solo operator whose business depends entirely on high-volume reasoning.
Downgrade when:
You realize you aren't using the reasoning models enough to justify the $2,400/year cost.
Switch vendor when:
You need to scale to multiple users (switch to Team).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Top-tier AI power user | $200 | $2,400 | Billed monthly. |
team standard
ChatGPT Team
$30/mo · $300/yr
Minimum 2 users required
Target users
small startups, creative agencies, departmental teams
Typical uses
- Collaborative prompt engineering
- Sharing custom GPTs within an organization
- Secure business data analysis
- Team-wide access to GPT-5
Why pick it
The best balance of data privacy, team collaboration, and high usage limits for small groups.
Key features
- Admin console for workspace management
- Data excluded from training by default
- Higher message caps than Plus
- Shared workspace for Custom GPTs
- Ability to bulk-manage member access
⚠ Marketing gimmicks to watch
The '2-User Minimum' Trap
The advertised $25/mo price is per user, but you cannot buy just one seat.
Impact: A solo user wanting 'Team' privacy must pay for two seats ($50/mo or $600/yr).
Annual Lock-in
The $25 rate requires paying for the full year upfront.
Impact: Monthly billing is 20% more expensive ($30/user/mo).
Pros
- Enterprise-grade privacy (no training on data)
- Higher usage limits than Plus
- Centralized billing for the team
Cons
- Minimum 2-seat requirement
- No SSO (Single Sign-On) at this tier
- Lacks the advanced security of Enterprise
Insider view
ChatGPT Team is the 'sweet spot' for most businesses. It provides the privacy that legal departments demand without the complex sales cycle of the Enterprise tier.
Max bang for buck
Pay annually to save $60 per user per year. For a 5-person team, that's $300 in savings.
🔒 Training-on-your-data policy
OpenAI does not train on ChatGPT Team data. Source: https://openai.com/chatgpt/team
🔄 Migration path
Upgrade when:
You need SSO, SCIM, or a dedicated account manager.
Downgrade when:
The team shrinks to one person (switch to Plus, but lose privacy).
Switch vendor when:
You need deep integration with Microsoft 365 (switch to Copilot for Business).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Small 2-person startup (Annual) | $50 | $600 | 2 users at $25/mo each, billed annually. |
| 5-person agency (Monthly) | $150 | $1,800 | 5 users at $30/mo each. |
team premium
Business ChatGPT & Codex
Per-token (see API rates above)
Optimized for technical teams
Target users
engineering teams, software houses, technical product managers
Typical uses
- Large-scale code generation and refactoring
- Technical documentation automation
- Internal tool development with Codex
Why pick it
A specialized tier that combines ChatGPT's conversational power with enhanced Codex capabilities for developers.
Key features
- Enhanced Codex model access
- Higher rate limits for technical queries
- Team collaboration tools
- Data privacy (no training on content)
- Priority support for technical issues
⚠ Marketing gimmicks to watch
Codex Deprecation Risk
OpenAI frequently updates its model lineup; 'Codex' features are increasingly being folded into standard GPT-5-pro.
Impact: Ensure the specific 'Codex' features you need aren't already available in cheaper tiers.
Volume-Based Pricing
While listed at $25, large teams may be pushed toward custom contracts.
Impact: Always ask for a volume discount if you are onboarding more than 50 developers.
Pros
- Superior coding performance
- Lower cost than standard Team tier for high-volume users
- Strong privacy protections
Cons
- Feature set can overlap with GitHub Copilot
- Requires annual commitment for best price
- Less focus on non-technical features
Insider view
This is essentially 'ChatGPT Team for Devs'. If your team spends 90% of their time in VS Code or terminal, this is the right tier. If they are doing marketing and sales, stick to the standard Team plan.
Max bang for buck
Integrate this with your internal CI/CD pipelines to automate code reviews and documentation.
🔒 Training-on-your-data policy
Data is not used for training. Source: https://openai.com/enterprise-privacy
🔄 Migration path
Upgrade when:
You need full Enterprise security features like SSO and BAA.
Downgrade when:
You find that standard GPT-5 is sufficient for your coding needs.
Switch vendor when:
You want the best-in-class IDE integration (switch to GitHub Copilot).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| 10-person dev team (Annual) | $200 | $2,400 | 10 users at $20/mo each. |
enterprise
ChatGPT Enterprise
Per-token (see API rates above)
Custom enterprise pricing — contact sales
Target users
fortune 500 companies, large government agencies, highly regulated industries
Typical uses
- Company-wide AI deployment
- Analyzing sensitive proprietary data
- Building custom internal AI applications
- HIPAA-compliant AI workflows
Why pick it
The only tier offering unlimited, high-speed access to frontier models with maximum security and administrative control.
Key features
- Unlimited, high-speed GPT-5 (no caps)
- SSO, SAML, and SCIM integration
- HIPAA and SOC2 compliance eligibility
- Advanced data analytics with unlimited usage
- Dedicated account management and support
- Shared templates and advanced admin controls
⚠ Marketing gimmicks to watch
The 'Unlimited' Premium
OpenAI charges a significant premium for 'unlimited' usage that many companies never actually hit.
Impact: Audit your Team-tier usage before upgrading to ensure the 'unlimited' benefit justifies the cost.
Negotiation Thresholds
Discounts typically only kick in at 150+ seats or multi-year commitments.
Impact: Smaller enterprises (50-100 seats) often pay the highest per-user rates.
Pros
- No message caps whatsoever
- Fastest response times (priority compute)
- Enterprise-grade security and compliance
Cons
- Expensive and requires a sales contract
- Can be overkill for smaller organizations
- Longer implementation time due to IT requirements
Insider view
Enterprise is about peace of mind. For a large company, the cost is secondary to the security (SSO) and the removal of productivity-killing message caps. It is the 'gold standard' for corporate AI.
Max bang for buck
Negotiate for 'Guaranteed Capacity' if you plan to build high-volume internal agents on top of the platform.
🔒 Training-on-your-data policy
Customer data is not used for training. OpenAI offers BAA for HIPAA compliance. Source: https://openai.com/enterprise-privacy
🔄 Migration path
Upgrade when:
You hit message caps on Team tier or require SSO for security compliance.
Downgrade when:
Usage drops significantly or budget cuts require moving to a per-user model.
Switch vendor when:
You require deep integration with the AWS ecosystem (switch to Bedrock/Claude).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| 200-user corporate deployment | varies | varies | Typically ranges from $30-$60 per user depending on negotiation. |
developer api
OpenAI API
Per-token (see API rates above)
Pay-as-you-go based on token usage
Target users
software developers, AI researchers, SaaS founders
Typical uses
- Integrating GPT-5 into third-party apps
- Automated content generation at scale
- Building custom AI agents
- Fine-tuning models on specific datasets
Why pick it
The most flexible and scalable way to build on OpenAI's technology with granular control over costs and model selection.
Key features
- Access to all models (GPT-5, o3, o4-mini)
- Prompt caching for 50% discount on repeat input
- Batch API for 50% discount on non-urgent tasks
- Fine-tuning capabilities
- Usage-based billing with tiered discounts
⚠ Marketing gimmicks to watch
Hidden Reasoning Token Billing
o-series models generate 'reasoning tokens' that are billed as output tokens but never appear in the response.
Impact: A single request can cost 5x more than expected if the model 'thinks' extensively.
High-Traffic Cache Routing Overflow
If you send the same prefix too fast (>15 RPM), OpenAI may route to a server without the cache, charging full price.
Impact: Use 'prompt_cache_key' to ensure consistent routing and avoid surprise costs.
Pros
- Only pay for what you use
- Access to the most powerful reasoning models (o1/o3)
- Highly reliable infrastructure
Cons
- Costs can spiral without strict monitoring
- Complex pricing (input vs output vs reasoning tokens)
- Rate limits can be restrictive for new accounts
Insider view
The API is where the real power lies, but it requires engineering discipline. The addition of prompt caching has made it much more affordable for RAG (Retrieval-Augmented Generation) workflows.
Max bang for buck
Use the Batch API for any task that can wait 24 hours; it's an instant 50% cost reduction.
🔒 Training-on-your-data policy
Data submitted via the API is not used to train OpenAI models. Source: https://openai.com/enterprise-privacy
🔄 Migration path
Upgrade when:
You need guaranteed throughput (switch to Provisioned Throughput).
Downgrade when:
You are only using it for personal tasks (switch to ChatGPT Plus).
Switch vendor when:
You need a 2-million token context window (switch to Gemini).
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Small app (1M tokens/mo GPT-5-mini) | $2.25 | $27 | Based on $0.25 input / $2.00 output per MTok. |
| Enterprise RAG (100M tokens/mo GPT-5) | $1125 | $13,500 | Assumes heavy use of prompt caching. |
consumer
ChatGPT Pro
$200/mo · $2400/yr
$200/mo
Target users
individual users, professionals
Key features
- Access to GPT-5.5 Pro
- Highest priority access
- Extended usage limits
| Scenario | Monthly | Annual | Notes |
|---|---|---|---|
| Standard annual cost | $200 | $2,400 | Monthly subscription |
All OpenAI products at a glance
Scroll up to the product profile for full detail
| Product | Price | Best for | Headline feature | Yearly estimate |
|---|---|---|---|---|
| ChatGPT Free | $0 | Casual search & chat | GPT-4o mini access | $0 |
| ChatGPT Go | Varies (Entry-paid) | Light users | Increased caps over Free | Varies |
| ChatGPT Plus | $20/mo | Individual power users | DALL-E & GPT-4o | $240 |
| ChatGPT Pro | $200/mo | Elite researchers | o1-pro mode | $2,400 |
| ChatGPT Team | $25-30/user/mo | Small collaborations | Admin workspace | $300-360/user |
| OpenAI API | Usage-based | App development | Token-based billing | Variable |
OpenAI vs the field
Same-tier comparison across top 5 vendors
| Comparison tier | Anthropic | OpenAI | xAI | Verdict | |
|---|---|---|---|---|---|
| Flagship Consumer ($20/mo) | Claude Pro $20/mo |
ChatGPT Plus $20/mo |
Gemini Advanced $20/mo |
Grok Premium $16/mo |
OpenAI offers the most robust multimodal toolset (DALL-E/Voice); Anthropic is often preferred for long-form writing. |
| High-End Consumer ($200/mo) | N/A N/A |
ChatGPT Pro $200/mo |
N/A N/A |
N/A N/A |
OpenAI currently stands alone in the $200 individual tier with its specialized o1-pro reasoning model. |
| SOTA API (Input/Output per 1M) | Claude 3.5 Sonnet $3 / $15 |
o1-preview $15 / $60 |
Gemini 1.5 Pro $1.25 / $10 |
— |
DeepSeek is the clear price leader; OpenAI o1 maintains a lead in complex reasoning benchmarks. |
| Standard Team Tier | Claude Team $25/mo (Annual) |
ChatGPT Team $25/mo (Annual) |
Gemini Business $20/mo (Annual) |
— |
Microsoft wins on Office integration; OpenAI wins on custom GPT ecosystem and ease of use. |
🌳 Which OpenAI product fits you?
3 questions, 1 recommendation
Are you building an application or using a chat interface?
Recommended
openai api
The API is the only way to programmatically access OpenAI models for custom software development.
See full profile ↑
Recommended
openai pro 200 + chatgpt pro
This tier is specifically built for users whose workflows are bottlenecked by standard reasoning model caps.
See full profile ↑
Recommended
openai plus + openai free
The standard choice for individuals who want the full suite of multimodal tools and higher caps than the free tier.
See full profile ↑
Recommended
openai enterprise
Necessary for large organizations requiring administrative control, security compliance, and unlimited high-speed access.
See full profile ↑
Recommended
openai team + openai business
The most cost-effective way for small groups to collaborate in a shared environment without training on business data.
See full profile ↑
How OpenAI pricing has moved
Tracking shifts in token rates and context billing thresholds.
→
GPT-5.4
— $2.5 / $15 per MTok now
No notable changes detected for gpt-5-4 in the recent changelog window.
→
GPT-5.4 Mini
— $0.75 / $4.5 per MTok now
No notable changes detected for gpt-5-4-mini in the recent changelog window.
→
GPT-5.4 Nano
— $0.2 / $1.25 per MTok now
No notable changes detected for gpt-5-4-nano in the recent changelog window.
→
GPT-5.4 Pro
— $30 / $180 per MTok now
No notable changes detected for gpt-5-4-pro in the recent changelog window.
→
GPT-5
— $1.25 / $10 per MTok now
No notable changes detected for gpt-5 in the recent changelog window.
→
GPT-5.5
— $5 / $30 per MTok now
No notable changes detected for gpt-5-5 in the recent changelog window.
→
GPT-5.5 Pro
— $30 / $180 per MTok now
No notable changes detected for gpt-5-5-pro in the recent changelog window.
→
GPT-5 Mini
— $0.25 / $2 per MTok now
No notable changes detected for gpt-5-mini in the recent changelog window.
→
GPT-5 Pro
— $15 / $120 per MTok now
No notable changes detected for gpt-5-pro in the recent changelog window.
→
GPT-5 Nano
— $0.05 / $0.4 per MTok now
No notable changes detected for gpt-5-nano in the recent changelog window.
→
o4-mini-2025-04-16
— $4 / $16 per MTok now
No notable changes detected for o4-mini-2025-04-16 in the recent changelog window.
→
o3-deep-research
— $5 / $20 per MTok now
No notable changes detected for o3-deep-research in the recent changelog window.
Overall: OpenAI is shifting away from simple per-token pricing toward more complex billing structures, including time-based metrics for Realtime APIs and tighter thresholds for long-context surcharges.
API or subscription: which is cheaper for you?
Cross-over math at current rates
openai-plus ($20/mo) vs gpt-5 API ($1.25/$10 per MTok)
Break-even: ~9,697 messages/month (avg ~600 tokens each)
At a 3:1 input/output ratio, each message costs approximately $0.00206 via API. You must send over 9,600 messages monthly to make the $20 subscription cheaper than pay-as-you-go API usage.
👉 API is significantly cheaper for casual users; Subscription is better for heavy daily chat users who value the UI and integrated tools.
chatgpt-pro ($200/mo) vs gpt-5-pro API ($15/$120 per MTok)
Break-even: ~8,081 messages/month (avg ~600 tokens each)
The Pro tier targets high-compute tasks. With GPT-5-Pro API rates at $15/$120, a single 600-token message costs $0.02475. The $200 subscription breaks even at roughly 8,000 messages.
👉 The subscription is a hedge against high-compute costs for power users; API is preferred only for low-volume, high-precision automation.
openai-go ($8/mo) vs gpt-5-mini API ($0.25/$2 per MTok)
Break-even: ~19,394 messages/month (avg ~600 tokens each)
GPT-5-mini is extremely efficient. At $0.0004125 per message, you would need to send nearly 20,000 messages a month to justify the $8 'Go' subscription on cost alone.
👉 Buy the subscription for the mobile app experience and convenience, not for cost savings over the API.
Rule of thumb
Subscriptions provide predictable costs and access to the ChatGPT UI/ecosystem, while the API offers granular control and often lower total costs for users sending fewer than 5,000 messages per month.
Current pricing (all production models)
| Model | Input $/M | Output $/M | Cached $/M | Context |
|---|---|---|---|---|
GPT-5.4gpt-5-4 |
$2.5 | $15 | $0.25 | 1,050,000 |
GPT-5.4 Minigpt-5-4-mini |
$0.75 | $4.5 | $0.075 | 400,000 |
GPT-5.4 Nanogpt-5-4-nano |
$0.20 | $1.25 | $0.020 | 272,000 |
GPT-5.4 Progpt-5-4-pro |
$30 | $180 | $3 | 1,050,000 |
GPT-5gpt-5 |
$1.25 | $10 | $0.13 | 400,000 |
GPT-5.5gpt-5-5 |
$5 | $30 | $0.50 | 1,050,000 |
GPT-5.5 Progpt-5-5-pro |
$30 | $180 | $3 | 1,050,000 |
GPT-5 Minigpt-5-mini |
$0.25 | $2 | $0.025 | 400,000 |
GPT-5 Progpt-5-pro |
$15 | $120 | — | 400,000 |
GPT-5 Nanogpt-5-nano |
$0.050 | $0.40 | $0.005 | 400,000 |
o4-mini-2025-04-16o4-mini-2025-04-16 |
$4 | $16 | $1 | — |
o3-deep-researcho3-deep-research |
$5 | $20 | — | — |
Pricing verified as of 2026-05-27. Prompt caching and Batch API (50% discount) available for most models.
Full rate breakdown (all variants)
Variants beyond standard API: batch (async, 50% off), cached read (0.1x), cache writes (1.25x or 2x base), long-context tier (~2x above threshold).
GPT-5.4 gpt-5-4
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $2.5 | $15 | Default per-token API rate |
| Batch API | $1.25 | $7.5 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $0.25 | $15 | Cached prompt input (~0.1x base); output rate unchanged |
| Long context (>270,000 tokens) | $5 | $22.5 | Higher rate applies above 270,000 tokens |
GPT-5.4 Mini gpt-5-4-mini
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $0.75 | $4.5 | Default per-token API rate |
| Batch API | $0.38 | $2.25 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $0.075 | $4.5 | Cached prompt input (~0.1x base); output rate unchanged |
GPT-5.4 Nano gpt-5-4-nano
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $0.20 | $1.25 | Default per-token API rate |
| Batch API | $0.10 | $0.63 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $0.020 | $1.25 | Cached prompt input (~0.1x base); output rate unchanged |
GPT-5.4 Pro gpt-5-4-pro
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $30 | $180 | Default per-token API rate |
| Batch API | $15 | $90 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $3 | $180 | Cached prompt input (~0.1x base); output rate unchanged |
| Long context (>270,000 tokens) | $60 | $270 | Higher rate applies above 270,000 tokens |
GPT-5 gpt-5
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $1.25 | $10 | Default per-token API rate |
| Batch API | $0.63 | $5 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $0.13 | $10 | Cached prompt input (~0.1x base); output rate unchanged |
GPT-5.5 gpt-5-5
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $5 | $30 | Default per-token API rate |
| Batch API | $2.5 | $15 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $0.50 | $30 | Cached prompt input (~0.1x base); output rate unchanged |
| Long context (>272,000 tokens) | $10 | $45 | Higher rate applies above 272,000 tokens |
GPT-5.5 Pro gpt-5-5-pro
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $30 | $180 | Default per-token API rate |
| Batch API | $15 | $90 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $3 | $180 | Cached prompt input (~0.1x base); output rate unchanged |
| Long context (>272,000 tokens) | $60 | $270 | Higher rate applies above 272,000 tokens |
GPT-5 Mini gpt-5-mini
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $0.25 | $2 | Default per-token API rate |
| Batch API | $0.13 | $1 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $0.025 | $2 | Cached prompt input (~0.1x base); output rate unchanged |
GPT-5 Pro gpt-5-pro
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $15 | $120 | Default per-token API rate |
| Batch API | $7.5 | $60 | Async batch processing, results within 24 hours, typically 50% off |
GPT-5 Nano gpt-5-nano
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $0.050 | $0.40 | Default per-token API rate |
| Batch API | $0.025 | $0.20 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $0.005 | $0.40 | Cached prompt input (~0.1x base); output rate unchanged |
o4-mini-2025-04-16 o4-mini-2025-04-16
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $4 | $16 | Default per-token API rate |
| Batch API | $2 | $8 | Async batch processing, results within 24 hours, typically 50% off |
| Cached read | $1 | $16 | Cached prompt input (~0.1x base); output rate unchanged |
o3-deep-research o3-deep-research
| Variant | Input $/M | Output $/M | Notes |
|---|---|---|---|
| Standard | $5 | $20 | Default per-token API rate |
Subscription plans (consumer + business)
| Plan | Audience | Monthly | Annual | Per seat | What's included |
|---|---|---|---|---|---|
| Business ChatGPT & Codex | business | $25 | $20/mo billed annually ($240/yr total) |
$1 |
Everything in ChatGPT Plus and Business Codex plans · Unlimited core chat and access to the best models for work · 60+ apps that bring your tools and data into ChatGPT — like Slack, Google Drive, SharePoint, GitHub, Atlassian, and more · Business features like apps, data analysis, record mode, canvas, shared projects, and custom workspace GPTs · Easy member, role, & billing management · A secure, dedicated workspace with essential admin controls, SAML SSO, and MFA · No training on your data; SAML security · Support for compliance with GDPR, CCPA, and other privacy laws. Aligned with CSA STAR and SOC 2 Type 2
Limits: min seats: 2 · limits source: transcript: 'Unlimited core chat'; specific quotas not published chatgpt.com ↗ |
| ChatGPT Enterprise Custom | enterprise | Contact | — | $1 |
Expanded context window that supports longer inputs and larger files · Enterprise-level security and controls, including SCIM, EKM, user analytics, domain verification, and role-based access controls · Advanced data privacy with custom data retention policies, encryption at rest and in transit, and no training on your business data by default · Support for data residency in ten regions · 24/7 priority support, SLAs, custom legal terms, and access to AI advisors (eligible customers) · Invoicing and billing, volume discounts
chatgpt.com ↗ |
| Business Codex | developer | — | — | — |
AI-powered software engineering · Automated code and security reviews · Automate tasks on your computer · Take action across your documents, tools, and codebases · Built-in worktrees and cloud environments for multi-agent workflows · No fixed seat fee; pay as you go based on usage · A secure, dedicated workspace with essential admin controls, SAML SSO, and MFA · No training on your data; SAML security · Support for compliance with GDPR, CCPA, and other privacy laws. Aligned with CSA STAR and SOC 2 Type 2
Limits: model: usage_based · limits source: no included quota; pure consumption chatgpt.com ↗ |
|
ChatGPT Free
ChatGPT Free |
consumer | $0 | $0 | — |
Access to GPT-5.5 Instant · Basic features
Limits: messages per 5h: 10 openai.com ↗ |
|
ChatGPT Go
ChatGPT |
consumer | $8 | — | — |
Ads · Notes: Has ads (US, India, others); launched India Aug 2025; rolled out 170+ countries
Limits: default model: gpt-5.3-instant · usage multiplier: 2x_free · deep research per month: 0 · codex context window tokens: 400000 openai.com ↗ |
|
ChatGPT Plus
ChatGPT |
consumer | $20 | — | — |
Access to GPT-5.5 · Advanced features · Priority access during peak times
Limits: context window: 128K tokens · messages per 5h: 100 openai.com ↗ |
|
ChatGPT Pro $100
ChatGPT |
consumer | $100 | — | — |
Notes: Launched April 9, 2026; 5x Plus quotas, 50 Deep Research/mo
Limits: default model: gpt-5.5-thinking · usage multiplier: 5x_plus · deep research per month: 50 openai.com ↗ |
|
ChatGPT Pro $200
ChatGPT |
consumer | $200 | — | — |
Notes: Top tier; 20x Plus quotas, 250 Deep Research/mo, GPT-5.4 1M context, full Sora
Limits: default model: gpt-5.4 · usage multiplier: 20x_plus · context window tokens: 1000000 · deep research per month: 250 openai.com ↗ |
|
ChatGPT Pro
ChatGPT |
consumer | $200 | — | — |
Access to GPT-5.5 Pro · Highest priority access · Extended usage limits
Limits: context window: 300K tokens · messages per 5h: 500 openai.com ↗ |
|
ChatGPT Team
ChatGPT |
team | $30 | $25/mo billed annually ($300/yr total) |
$1 |
Shared workspace · Admin controls · Team billing
Limits: seats: 2-150 · context window: 128K tokens · messages per 5h: 100 openai.com ↗ |
Subscription pricing is separate from per-token API rates above.
What changed in the last 30-90 days
- 2026-05-26: GPT-5.5 Pro and GPT-5.5 adjusted longContextThreshold from 272,000 to 270,000. — Slight adjustment to when long-context pricing or processing rules apply.
- 2026-05-26: GPT-Realtime 2 cost per minute established at $0.034. — Provides a clear benchmark for real-time audio and multimodal interaction costs.
- 2026-05-24: New models added: GPT-Realtime Whisper, GPT-Image 2, GPT-Realtime 2, and GPT-Realtime Translate. — Expands the multimodal capabilities available via API for specialized audio and image tasks.
- 2026-05-19: GPT-5.4 Pro and GPT-5.4 longContextThreshold reduced from 270,000 to 200,000. — Buyers will hit long-context processing thresholds earlier in their token usage.
How buyers think about OpenAI pricing
Each scenario below is interactive — tweak the inputs to see how the math changes for your workload.
Cheapest GPT tier for high-volume classification
The problem: You need to process hundreds of thousands of simple classification tasks, such as sentiment analysis or lead scoring, without exhausting your monthly budget on expensive frontier models.
What to do: Use GPT-5 Nano for ultra-low-cost processing of simple, high-volume tasks.
Processing 1,000,000 input tokens costs $0.05 and 1,000,000 output tokens costs $0.40. For a batch of 100,000 classification tasks with 500 input tokens and 50 output tokens each, the total cost is $2.50 for input and $2.00 for output, totaling $4.50 (as of 2026-05-27).
→ GPT-5 Nano processes two million tokens for under $0.50.
Is ChatGPT Plus at $20 per month worth it versus paying API
The problem: You are trying to decide if a fixed $20 monthly subscription for ChatGPT Plus is more economical than paying for metered API usage for your daily coding and research tasks.
What to do: Compare your monthly token volume against the GPT-5 API rates to find your break-even point.
At GPT-5 rates of $1.25 per million input tokens, a $20 monthly fee equals 16,000,000 input tokens. If your monthly usage exceeds 16 million input tokens, or a proportional mix of input and output tokens, the $20 Plus subscription is more cost-effective (as of 2026-05-27).
→ High-volume users save more by sticking to the $20 Plus subscription.
When GPT-5.5 Pro is worth the premium
The problem: You are unsure if the significant price jump to the Pro tier is justified for your specific enterprise workflows or if the standard model suffices.
What to do: Reserve GPT-5.5 Pro for high-stakes complex reasoning while using GPT-5.5 for standard premium work.
A request with 1,000 input tokens and 1,000 output tokens costs $0.035 on GPT-5.5 ($5 per million input and $30 per million output). The same request on GPT-5.5 Pro costs $0.21 ($30 per million input and $180 per million output), representing a 6x price increase (as of 2026-05-27).
→ GPT-5.5 Pro costs 6x more than the standard GPT-5.5 for high-stakes reasoning.
When o4-mini reasoning beats frontier chat models
The problem: You need deep logical reasoning for technical tasks but want to avoid the high costs associated with flagship frontier models.
What to do: Deploy o4-mini-2025-04-16 for medium-stakes reasoning tasks that require more than standard chat capabilities.
A request using 1,000 input tokens and 1,000 output tokens on o4-mini costs $0.02 ($4 per million input and $16 per million output). This provides specialized reasoning capabilities at a predictable rate for complex logic (as of 2026-05-27).
→ o4-mini provides specialized reasoning at a competitive rate compared to frontier models.
Cutting cost 50 percent with the Batch API
The problem: You have large-scale workloads like data enrichment or document summarization that do not require immediate real-time responses.
What to do: Route all 24-hour-tolerant workloads through the Batch API endpoint to receive an automatic 50 percent discount.
Processing 1,000,000 input tokens on GPT-5.4 costs $2.50 via the standard API. Using the Batch API for the same volume reduces the cost to $1.25, providing immediate savings for asynchronous tasks (as of 2026-05-27).
→ Batch API provides a flat 50 percent discount for non-urgent processing.
When prompt caching pays off
The problem: Your AI agents use long system prompts or large context windows repeatedly, leading to high redundant input costs.
What to do: Leverage automatic prompt caching for prefixes longer than 1,024 tokens to reduce input expenses.
On GPT-5.4, standard input tokens cost $2.50 per million. Cached input tokens cost only $0.25 per million. If an agent processes 1,000,000 tokens with a 100 percent cache hit rate, you save $2.25 per million tokens (as of 2026-05-27).
→ Prompt caching reduces input costs by 90 percent for repeated prefixes.
Volume discounts & partner programs
Heads up — these are community-sourced and analyst-reported terms. Specific credit amounts, discount percentages, and program thresholds change frequently. Always verify current terms directly with OpenAI before relying on a specific number. Treat reported figures as ballpark, not contract language.
OpenAI Frontier Alliances
Threshold: Restricted to major Global Systems Integrators (GSIs) and advisory firms
Typical discount (reported): varies by contract
Benefits:
- Access to Forward Deployed Engineers (FDEs)
- Technical resources and roadmap insight
- Certification on OpenAI technology
- Priority access to product and research teams
How to engage: Direct partnership with OpenAI GTM Partnerships team; current partners include BCG, McKinsey, Accenture, and Capgemini
Source: constellationr.comanalyst_report · cited 2026-02-24
OpenAI for Startups / VC Partner Program
Threshold: Requires referral from a partner Venture Capital firm
Typical discount (reported): up to $100,000 in free API credits
Benefits:
- Free API credits for GPT-5.2 and other models
- Usage tier upgrades
- Exclusive access to events and resources
- Access to new agent infrastructure
How to engage: Apply through a participating VC firm or via the OpenAI VC Partner application page
Source: openai.comvendor_official · cited 2026-05-27
OpenAI Guaranteed Capacity
Threshold: Enterprise-scale commitments (up to 1 billion tokens per minute)
Typical discount (reported): varies by commitment length (1, 2, or 3 years)
Benefits:
- Guaranteed compute resources for AI services
- Stable supply for AI agents and high-volume workflows
- Protection against capacity-constrained environments
How to engage: Direct negotiation with OpenAI enterprise sales; available on a first-come, first-served basis
Source: eweek.comanalyst_report · cited 2026-05-22
Azure OpenAI Provisioned Throughput Units (PTU)
Threshold: Typically 50 to 100 PTU minimum for flagship models
Typical discount (reported): reportedly 18% to 34% below pay-as-you-go
Benefits:
- Predictable monthly or annual costs
- Reserved capacity with no queuing or 429 throttling
- Integration with Azure MACC (Microsoft Azure Consumption Commitment)
- Regional data residency and private networking
How to engage: Purchase through Azure Portal or via Microsoft Enterprise Agreement (EA)
Source: redresscompliance.comcommunity · cited 2026-02-08
ChatGPT Enterprise Volume Negotiation
Threshold: Typically requires a minimum of ~150 users
Typical discount (reported): up to 20% for multi-year contracts
Benefits:
- Uncapped usage and no message rate limits
- HIPAA Business Associate Agreement (BAA) eligibility
- 24/7 priority support and enterprise-grade SLAs
- Extended context windows for frontier models
How to engage: Contact OpenAI enterprise sales team
Source: moomoo.comanalyst_report · cited 2025-06-18
YC Startup Batch Credits
Threshold: Acceptance into Y Combinator (e.g., S25 or S26 batches)
Typical discount (reported): $2 million in API credits
Benefits:
- Prepaid compute for model training and inference
- Structured as an uncapped SAFE
- Deep visibility into OpenAI roadmap
How to engage: Apply and be accepted into the Y Combinator startup accelerator
Source: reddit.comcommunity · cited 2026-05-22
Multi-cloud availability
Cloud-marketplace terms change frequently. Model availability dates, pricing parity, and regional features can drift week to week. Verify with each cloud's pricing page (AWS Bedrock, Google Vertex, Azure AI Foundry) before architecting around specifics.
| Cloud | Model availability | Price vs vendor-direct | Reasons to pick |
|---|---|---|---|
| Microsoft Azure (Azure OpenAI Service) | GPT-5, GPT-5.5, GPT-5.4, GPT-4.1, o4-mini, o3, GPT-image-1.5, and gpt-4o audio models | matches this almost exactly for pay-as-you-go; Provisioned Throughput Units (PTUs) available for reserved capacity |
vertexaisearch.cloud.google.com ↗ |
| AWS Bedrock | GPT-5.5, GPT-5.4, Codex, and Bedrock Managed Agents (powered by OpenAI) | reportedly per-token parity, but with 15-40% in infrastructure overhead (VPC endpoints, CloudWatch, etc.) |
vertexaisearch.cloud.google.com ↗ |
| Google Vertex AI (Gemini Enterprise Agent Platform) | GPT-5.3 Instant (preview) | matches the rates that the API endpoint originally inherited |
vertexaisearch.cloud.google.com ↗ |
| Together.ai | gpt-oss-120b and gpt-oss-20b (open weights models) | transparent pricing for open models; up to $50,000 in credits for qualifying startups |
vertexaisearch.cloud.google.com ↗ |
| Anyscale | gpt-oss-120b | reportedly $0.10-$0.30 per million tokens for open-weight models; cited 6.1x cost savings on LLM inference |
vertexaisearch.cloud.google.com ↗ |
Free credits & startup programs
Program details and credit amounts shift often. Apply directly through each program's official page for current values, eligibility windows, and application requirements.
OpenAI & Y Combinator Partnership ($2M Deal)
Reported value: $2 million in API credits
Eligibility: Every startup in the Y Combinator Spring 2026 (S25) and Summer 2026 (S26) batches.
How to apply: Automatic for accepted YC batch companies; requires signing an uncapped SAFE (Simple Agreement for Future Equity) with OpenAI.
OpenAI Researcher Access Program
Reported value: up to $1,000 in API credits
Eligibility: Researchers with active affiliation to an academic institution, research organization, or nonprofit conducting research on AI safety, alignment, or societal impact.
How to apply: Apply via the official OpenAI Researcher Access Program portal (hosted on SurveyMonkey Apply); applications reviewed quarterly in March, June, September, and December.
Microsoft for Startups Founders Hub
Reported value: up to $150,000 in Azure credits and $2,500 in OpenAI API credits
Eligibility: Early-stage startups; higher tiers ($100k-$150k) typically require affiliation with the Microsoft for Startups Investor Network.
How to apply: Apply at microsoft.com/startups; basic tier ($1,000-$5,000) is often instant approval for verified businesses.
OpenAI Startup Program (Tiered)
Reported value: reportedly $2,500 (Tier 1) to $100,000+ (Tier 3)
Eligibility: Tier 1 is reportedly self-serve for eligible startups; Tier 2 and 3 require referral codes from partner VCs or accelerators.
How to apply: Apply at openai.com/startups; Tier 2+ requires a partner-provided referral code (format typically PARTNER-XXXX-XXXX).
Ramp / Brex OpenAI Startup Perk
Reported value: up to $2,500 in OpenAI API credits
Eligibility: Startups using Ramp or Brex for corporate banking/spend management.
How to apply: Claim via the 'Perks' or 'Rewards' dashboard within the Ramp or Brex platform.
OpenAI Safety Fellowship
Reported value: stipends and access to OpenAI models
Eligibility: External researchers, engineers, and practitioners studying AI risks (robustness, privacy, misuse prevention).
How to apply: Six-month program running from September 2026 to February 2027; requires application detailing research proposal.
OpenAI Grove
Reported value: early access to new tools and models
Eligibility: Pre-idea individuals and technical talent at the start of their company-building journey.
How to apply: Application-based cohort program (approximately 15 participants); includes 5 weeks of programming at OpenAI HQ.
Pricing gotchas to watch
Most gotchas below were surfaced by community reports. Some may have been fixed, changed, or never been the user-facing issue they appeared. Verify against current vendor docs before architecting around a workaround.
High-Traffic Cache Routing Overflow
OpenAI's automatic prompt caching routes requests based on a hash of the first ~256 tokens. In high-traffic scenarios exceeding approximately 15 requests per minute for the same prefix, traffic can overflow to additional servers that do not hold the cache, resulting in unexpected full-price charges despite identical prefixes.
Workaround: Use the optional 'prompt_cache_key' parameter to influence routing and improve cache hit consistency for shared prefixes.
Source: medium.comblog_post · cited 2026-05-10
Hidden Reasoning Token Billing
Models in the o-series (o1, o3, o4-mini) and GPT-5 generate internal 'reasoning tokens' that are not visible in the final API response but are billed as output tokens. These tokens can reportedly cost up to 5x the standard output rate and significantly inflate the total cost of a single request.
Workaround: Set the 'max_completion_tokens' parameter to place a hard cap on the total tokens generated, which includes both visible output and hidden reasoning tokens.
Source: reddit.comreddit · cited 2025-08-27
Parallel Tool Call Token Overhead
Enabling 'parallel_tool_calls' (which is true by default) reportedly consumes approximately 199 tokens of overhead. Additionally, simply including any tools in a request adds a fixed overhead of 16 tokens (3 for the system message plus the template).
Workaround: Set 'parallel_tool_calls' to false if simultaneous tool execution is not required to save on per-request token overhead.
Source: github.comgithub_issue · cited 2024-10-26
Azure Regional Pricing Variance
Azure OpenAI pricing is not globally uniform. While North American rates are the baseline, some regions like Brazil South reportedly carry a +60% premium, while others like Central India may offer a -30% discount. Data egress costs also vary by region, ranging from approximately $0.04 to $0.087 per GB.
Workaround: Use the Azure Pricing Calculator to verify region-specific rates and consider deploying in lower-cost regions if data residency requirements allow.
Source: checkthat.aiblog_post · cited 2026-03-30
Prompt Cache TTL and Eviction Limits
In-memory prompt caching typically persists for only 5–10 minutes of inactivity, with a hard eviction ceiling of one hour regardless of traffic volume. This can lead to frequent cache misses for applications with sparse or irregular traffic patterns.
Workaround: For supported models (e.g., GPT-5.4, GPT-5.5), use 'Extended' prompt cache retention to increase the TTL up to a maximum of 24 hours.
Source: openai.comvendor_docs · cited 2026-02-18
High-Detail Image Tiling Costs
When using vision models in 'high' detail mode, images are divided into 512x512 pixel tiles. Each tile costs 170 tokens, plus a base overhead of 85 tokens. A large image (e.g., 1024x1024) can quickly scale to 765 tokens (4 tiles + base), whereas 'low' detail mode is a fixed 85 tokens regardless of size.
Workaround: Use 'detail: low' for tasks that do not require fine-grained visual analysis to maintain a predictable 85-token cost per image.
Source: platform.openai.comvendor_docs · cited 2025-08-05
Hidden costs (25-40% beyond per-token rates)
- Hidden reasoning tokens in o-series and GPT-5 can cost up to 5x the standard output rate.
- Parallel tool call overhead adds approximately 199 tokens to every request when enabled.
- High-detail image tiling costs 170 tokens per 512x512 tile plus a base overhead of 85 tokens.
- Azure regional pricing variance can add a 60 percent premium in regions like Brazil South.
- Cache routing overflow can result in full-price charges if traffic exceeds 15 requests per minute for the same prefix.
- Prompt cache eviction occurs after 5 to 10 minutes of inactivity, leading to unexpected misses for sparse traffic.
- Data egress and private networking fees on cloud providers add 5 to 15 percent to the total infrastructure bill.
Typical overhead: 25-40% beyond raw per-token rates.
What it costs to leave OpenAI
Migrating away from OpenAI involves rewriting prompts specifically tuned for GPT's instruction-following style and replacing proprietary features like Assistants API threads. While OpenAI-compatible wrappers exist, differences in reasoning token handling and tool-calling schemas require significant testing.
- small project (1-5 prompts): 1-2 engineer-days
- mid-size (10-50 prompts): 1-2 engineer-weeks
- large agentic system: 1-3 engineer-months
Who is this for?
For vibe coders & solo devs
For rapid prototyping, focus on the Nano model series to keep experimentation costs near zero. You can leverage startup perks like the $2,500 in API credits available through Ramp or Brex to fund your initial development. Use GPT-5 Nano for basic logic and only step up to GPT-5.4 when your application requires higher creative nuance.* Start with GPT-5 Nano at $0.05 per million input tokens.
* Claim the $2,500 credit perk if you use Ramp or Brex.
* Use the Batch API for non-interactive testing to double your credit runway.
* Monitor usage tiers to unlock higher rate limits as you scale.
For SMBs and growing teams
Small businesses should prioritize the ChatGPT Team plan for internal tools to gain higher message caps and administrative controls. For customer-facing apps, implement prompt caching immediately to handle repetitive user queries efficiently. This approach balances the predictable cost of seats with the scalability of the API.* Use the Team plan for internal staff to avoid per-token costs for research.
* Implement 'detail: low' for vision tasks to maintain a fixed 85-token cost per image.
* Set hard monthly billing limits in the OpenAI dashboard to prevent overages.
* Apply for the Microsoft for Startups Founders Hub to access up to $2,500 in OpenAI credits.
For enterprise buyers
Enterprise buyers should look toward Azure OpenAI for Provisioned Throughput Units (PTU) to secure predictable costs and 18 percent to 34 percent discounts. For massive scale, the Guaranteed Capacity program supports up to 1 billion tokens per minute. Engaging with the Frontier Alliances program can provide direct access to engineers for complex deployments.* Negotiate multi-year contracts for ChatGPT Enterprise to secure up to 20 percent discounts.
* Use Azure PTUs to bypass rate limits and ensure stable latency.
* Explore the Frontier Alliances for priority access to product roadmaps.
* Utilize the $2 million credit deal if your portfolio companies are in the Y Combinator S25 or S26 batches.
Need help deciding which OpenAI tier or model fits your workload?
Book a $19.99 quick call →
Sources verified for this page
Primary: OpenAI pricing page
View all 24 cited insider sources across 16 domains
- OpenAI Frontier Alliances (analyst_report, verified 2026-02-24)
- OpenAI for Startups / VC Partner Program (vendor_official, verified 2026-05-27)
- OpenAI Guaranteed Capacity (analyst_report, verified 2026-05-22)
- Azure OpenAI Provisioned Throughput Units (PTU) (community, verified 2026-02-08)
- ChatGPT Enterprise Volume Negotiation (analyst_report, verified 2025-06-18)
- YC Startup Batch Credits (community, verified 2026-05-22)
- High-Traffic Cache Routing Overflow (blog_post, verified 2026-05-10)
- Hidden Reasoning Token Billing (reddit, verified 2025-08-27)
- Parallel Tool Call Token Overhead (github_issue, verified 2024-10-26)
- Azure Regional Pricing Variance (blog_post, verified 2026-03-30)
- Prompt Cache TTL and Eviction Limits (vendor_docs, verified 2026-02-18)
- High-Detail Image Tiling Costs (vendor_docs, verified 2025-08-05)
- Microsoft Azure (Azure OpenAI Service) (grounded_research, verified 2026-05-13)
- AWS Bedrock (grounded_research, verified 2026-05-06)
- Google Vertex AI (Gemini Enterprise Agent Platform) (grounded_research, verified 2026-03-06)
- Together.ai (grounded_research, verified 2026-05-08)
- Anyscale (grounded_research, verified 2026-05-20)
- OpenAI & Y Combinator Partnership ($2M Deal) (grounded_research, verified 2026-05-26)
- OpenAI Researcher Access Program (grounded_research, verified 2026-04-17)
- Microsoft for Startups Founders Hub (grounded_research, verified 2026-04-14)
- OpenAI Startup Program (Tiered) (grounded_research, verified 2026-02-05)
- Ramp / Brex OpenAI Startup Perk (grounded_research, verified 2026-05-02)
- OpenAI Safety Fellowship (grounded_research, verified 2026-04-16)
- OpenAI Grove (grounded_research, verified 2026-01-02)
Generator: gen-v5.0.8-2026-05-25 · Last refreshed: Tue May 26 2026 22:55:07 GMT-0400 (Eastern Daylight Time) · Pricing snapshot: Tue May 26 2026 00:00:00 GMT-0400 (Eastern Daylight Time)