On this page
At a glance
| Vendor | How you move up | Top published tier | Monthly cap there | How to get more |
|---|---|---|---|---|
| OpenAI | Automatic, on total credit purchases | Grow at $500 purchased | $200,000 a month | Upgrade tier button. For enterprises, Scale Tier or Reserved Tier |
| Anthropic | Automatic, on usage history and account standing | Scale, no published threshold | $200,000 a month | Request tier increase in the Console, then sales |
| Google Gemini API | Automatic, on spend plus days since first payment | Tier 3 at $1,000 paid and 30 days | $20,000 to $100,000+ a month | Request form, no guarantee |
| xAI | Automatic, on spend since January 1, 2026 | Tier 4 at $5,000 spent | None listed | Console request, or Enterprise through sales |
| Mistral | No published tiers | None listed | Set by your organization | Contact support with your targets |
| Azure OpenAI | Automatic, on consumption and your Microsoft relationship | Tier 6, no published threshold | None listed | Quota request form |
| Amazon Bedrock | No tiers, quotas per model and Region | None listed | None listed | Service Quotas request, or an AWS Support case |
| Cohere | Switch from a trial key to a production key | Production key | None listed | Email support or sales |
| Groq | Upgrade to the Developer plan | Developer plan | None listed | Higher limits for select workloads |
| Fireworks AI | Automatic, on credits added or spend | Tier 4 at $5,000 | Prepaid accounts set their own limit | Contact Fireworks for Unlimited |
| Together AI | No published tiers | None listed | None listed | Provisioned throughput through sales |
| Perplexity | Automatic, on credits purchased | Tier 5 at $5,000 purchased | None listed | Rate limit increase form |
| Cerebras | First credit purchase moves you to Developer | Developer, pay as you go | None listed | Contact Cerebras for Enterprise limits |
OpenAI
Your tier follows how much you've bought in credits. It's set per organization, and limits also vary by model.
| Tier | Total credit purchases | Monthly usage limit | Astra, Sol, Terra RPM | Astra, Sol, Terra TPM |
|---|---|---|---|---|
| Free | Allowed geography | $100 | None listed | None listed |
| Build | $5 | $500 | 5,000 | 1,000,000 |
| Launch | $100 | $5,000 | 10,000 | 4,000,000 |
| Grow | $500 | $200,000 | 15,000 | 40,000,000 |
- If traffic keeps hitting ramp-rate limits, OpenAI points enterprise customers at Scale Tier, or Reserved Tier for GPT-5.6 and later. Source
How to raise your OpenAI limits
Anthropic
Anthropic doesn't publish what it takes to reach each tier. New organizations can start in a lower Evaluation tier.
| Tier | Monthly spend cap | Opus 5.5 RPM | Opus 5.5 input TPM | Opus 5.5 output TPM |
|---|---|---|---|---|
| Start | $500 | 4,000 | 2,000,000 | 400,000 |
| Build | $1,000 | 10,000 | 5,000,000 | 1,000,000 |
| Scale | $200,000 | 13,000 | 10,000,000 | 2,000,000 |
| Custom | None | Set with sales | Set with sales | Set with sales |
- Hit the spend cap and usage pauses until the first of next month, unless you get a higher limit sooner. Source
- On most Claude models, tokens read from the prompt cache don't count toward input TPM. Source
- New Priority Tier commitments are no longer sold. For guaranteed capacity, Anthropic says to contact sales. Source
How to raise your Anthropic limits
Google Gemini API
Tiers 2 and 3 count all your Google Cloud spend on the linked billing account, not only Gemini. Google can also limit spend over a rolling 10 minutes, depending on your billing history and account standing.
| Tier | You get there at | Monthly cap | Spend limit per 10 minutes |
|---|---|---|---|
| Free | Active project or free trial | None listed | None listed |
| Tier 1 | Linked active billing account | $250 | $10 |
| Tier 2 | $100 paid, 3 days after first payment | $2,000 | $50 |
| Tier 3 | $1,000 paid, 30 days after first payment | $20,000 to $100,000+ | $200 |
- Reach the monthly cap and service pauses for every project on that billing account until the next billing cycle. Source
How to raise your Google Gemini API limits
xAI
xAI lists per-second requests, worked out from a per-minute budget. Once you reach a tier you keep it.
| Tier | Cumulative spend | grok-4.7 RPS | grok-4.7 TPM |
|---|---|---|---|
| Tier 0 | $0 | 150 | 50M |
| Tier 1 | $50 | 172 | 53M |
| Tier 2 | $250 | 208 | 60M |
| Tier 3 | $1,000 | 312 | 74M |
| Tier 4 | $5,000 | 500 | 100M |
- Cached prompt tokens still count toward TPM on xAI, even though they're billed at a lower rate. Source
Mistral
Mistral doesn't publish limit numbers in its docs. Yours are in the Admin Panel under API, then Limits, listed per model as tokens per minute and requests per second.
- If the organization reaches its monthly spending limit, API access can be suspended until the next month or until an admin raises it. Source
From Mistral docs, Usage and limits, checked .
How to raise your Mistral limits
Azure OpenAI
There are seven quota tiers, set per Azure subscription. Enterprise Agreement customers are assigned higher ones.
| Tier | gpt-6-sol Global Standard RPM | gpt-6-sol Global Standard TPM |
|---|---|---|
| Tier 1 | 250 | 250,000 |
| Tier 2 | 1,000 | 1,000,000 |
| Tier 3 | 2,500 | 2,500,000 |
| Tier 4 | 7,000 | 7,000,000 |
| Tier 5 | 10,000 | 10,000,000 |
| Tier 6 | 15,000 | 15,000,000 |
- You can opt out of automatic upgrades if you use quota to control billing. Microsoft marks this as a preview that may change. Source
How to raise your Azure OpenAI limits
Amazon Bedrock
Bedrock has no tiers. Quotas are set per model and per Region, and the real values for your account can sit below the defaults. Check Service Quotas.
- Increases go first to customers whose traffic consumes their existing quota, and a request can be denied otherwise. Source
- AWS doesn't grant increases for models in Legacy or Deprecated status. Source
- The two endpoints, bedrock-runtime and bedrock-mantle, track quotas separately, and mantle increases go through an AWS Support form. Source
From Amazon Bedrock User Guide, Quotas, checked .
How to raise your Amazon Bedrock limits
Cohere
Cohere has two kinds of key, and newer model variants need a conversation with sales.
| Model | Trial key | Production key |
|---|---|---|
| Command A | 20 req/min | 500 req/min |
| Command A Reasoning | 20 req/min | Contact sales |
- Trial keys are limited to 1,000 API calls a month. Source
Groq
Groq publishes separate limits for its Free and Developer plans. Your exact numbers are on the Limits page in its console.
Fireworks AI
Serverless limits adapt to your traffic within a ceiling set by your spending tier and the model's size. The cap column only applies to older postpaid accounts.
| Tier | You get there at | Legacy postpaid monthly cap |
|---|---|---|
| Tier 1 | Valid payment method and billing profile | $50 |
| Tier 2 | $50 in spend or credits | $500 |
| Tier 3 | $500 in spend or credits | $5,000 |
| Tier 4 | $5,000 in spend or credits | $50,000 |
| Unlimited | Contact Fireworks | Unlimited |
- All API usage on an account shares one ceiling of 6,000 requests a minute, or 10 with no payment method or no credits. Source
Together AI
Together doesn't publish tier numbers. Serverless is best effort, and it may limit requests when demand is high.
- When a model is in especially high demand, Together may apply custom limits to that model. Source
From Together AI docs, Rate limits, checked .
Perplexity
Tiers follow total credits purchased, and your tier upgrades once you cross the threshold.
Cerebras
The free trial gives you $5 in credits that expire after 30 days. Your first purchase removes the hourly and daily token caps.
| Tier | gpt-oss-120b RPM | Uncached TPM | Total TPM |
|---|---|---|---|
| Free Trial | 5 | 30K | 90K |
| Developer | 1K | 1M | 3M |
| Enterprise | Set per organization | Set per organization | Set per organization |
Questions
Which AI vendors publish what it takes to reach each tier?
OpenAI, Google, xAI, Perplexity and Fireworks publish spend thresholds. Anthropic and Azure OpenAI have tiers but don't publish what it takes to reach them. Mistral, Together AI and Amazon Bedrock have no published tiers.
Does spending more raise your rate limits?
At most vendors, yes, and often on its own. OpenAI, xAI, Perplexity and Fireworks move you up on money spent or credits bought. Google adds a waiting period. Bedrock is the exception, since you request increases there and AWS looks at how much of your quota you already use.
Why don't these numbers match my console?
Your console wins. Vendors change limits per account, and Bedrock says your quota can be lower than its defaults. The tables here show one model per vendor, so your model may differ.
Will a second API key give me more rate limit?
Not at the vendors that say where limits sit. OpenAI sets them per organization and project, Anthropic per organization, Google per project and xAI per team.
Changelog
- November 2, 2026. First version, 13 vendors. Anthropic no longer sells new Priority Tier commitments. OpenAI's paid tiers are Build, Launch and Grow, so older tier numbers won't match.
Sources
- OpenAI API docs, Rate limits
- Claude API docs, Rate limits
- Claude API docs, Service tiers
- Gemini API docs, Rate limits
- Gemini API docs, Billing
- xAI docs, Rate limits
- Mistral docs, Usage and limits
- Microsoft Learn, Azure OpenAI quotas and limits
- Amazon Bedrock User Guide, Quotas
- Amazon Bedrock User Guide, Quotas for the bedrock-runtime endpoint
- Amazon Bedrock User Guide, Quotas for the bedrock-mantle endpoint
- Cohere docs, API keys and rate limits
- Groq docs, Rate limits
- Fireworks docs, Account quotas
- Together AI docs, Rate limits
- Perplexity docs, Rate limits and usage tiers
- Cerebras docs, Rate limits
Checked . Vendors change limits often, so the linked page is the final word.
The ernd card is a credit card for US companies, with its own partners in the AI stack. ernd isn't affiliated with any of the vendors on this page. ernd is run by Provenance Research Inc.