Published rate limits of 8 AI API providers: requests and tokens per minute by tier where the provider prints them, how a tier is reached, what counts toward the limit and what batch processing changes, in the providers’ own words, read .
Anthropic Claude API rate limits
Limits by model and tier
Model
Tier
Requests per minute
Input tokens per minute
Output tokens per minute
Claude Opus 5.5
Start
1,000
2,000,000
400,000
Claude Sonnet 5.5
Start
1,000
2,000,000
400,000
Claude Haiku 5.5
Start
1,000
2,000,000
400,000
Claude Opus 5.5
Build
5,000
5,000,000
1,000,000
Claude Sonnet 5.5
Build
5,000
5,000,000
1,000,000
Claude Haiku 5.5
Build
5,000
5,000,000
1,000,000
Claude Opus 5.5
Scale
10,000
10,000,000
2,000,000
Claude Sonnet 5.5
Scale
10,000
10,000,000
2,000,000
Claude Haiku 5.5
Scale
10,000
10,000,000
2,000,000
Tiers
Tier
Monthly spend cap, USD
Start
500
Build
1,000
Scale
200,000
Batch limits
Tier
Batch requests per minute
Requests waiting in batches
Requests per batch
Start
1,000
200,000
—
Build
2,000
300,000
—
Scale
4,000
500,000
—
All tiers
—
—
100,000
How the tier is set: “Limits are defined by usage tier. Organizations are placed on a tier automatically based on usage history and account standing and can move to a higher tier over time as they use the API.”
How the tier is set: “New organizations and organizations with limited usage history may start in the Evaluation tier, with limits below the standard limits shown on this page while account history is established.”
Spend cap: “Organizations on the Custom tier have no monthly spend cap; limits are arranged with their account team.”
How limits refill: “The API uses the token bucket algorithm to do rate limiting. This means that your capacity is continuously replenished up to your maximum limit, rather than being reset at fixed intervals.”
Cache and limits: “For most Claude models, only uncached input tokens count toward your ITPM rate limits.”
When a limit is hit: “If you exceed any of the rate limits you will get a 429 error describing which rate limit was exceeded. When the response has a retry-after header, wait that many seconds before retrying.”
Batch price: “The Batches API offers significant cost savings. All usage is charged at 50% of the standard API prices.”
Batch window: “Batches expire if processing does not complete within 24 hours.”
Batch timing: “The system processes each batch as fast as possible, with most batches completing within 1 hour.”
How the tier is set: “The three paid usage tiers are Build, Launch, and Grow. Your organization's usage tier upgrades automatically as its total credit purchases reach each threshold.”
Shared limits: “Some model families have shared rate limits.”
Long context: “For long context models like GPT-5.5, there is a separate rate limit for long context requests.”
Ramp-up: “As a rule of thumb, once your traffic reaches 1 million input tokens per minute (TPM), increase it by no more than 50% every 15 minutes.”
When a limit is hit: “When a request exceeds a temporary rate limit, the API returns a 429 error. The response can include a Retry-After header that tells you how many seconds to wait before trying again.”
Batch window: “For now, the completion window can only be set to 24h.”
Batch timing: “Fast completion times: Each batch completes within 24 hours (and often more quickly)”
Batch queue: “Queued prompt tokens per model: Each model has a maximum number of prompt tokens that can be queued for batch processing. You can find these limits on the Platform Settings page.”
How the tier is set: “Free Active project or free trial N/A”
How the tier is set: “Tier 3 Paid $1,000 + 30 days from first successful payment $20,000 - $100,000+”
Spend window: “These limits are evaluated on a rolling 10-minute window.”
What is limited: “Requests per minute ( RPM ) Tokens per minute (input) ( TPM ) Requests per day ( RPD )”
Who the limit applies to: “Rate limits are applied per project, not per API key. Requests per day ( RPD ) quotas reset at midnight Pacific time.”
Where the numbers are: “Rate limits depend on a variety of factors (such as your usage tier) and can be viewed in Google AI Studio. As your tier and account status change over time, your rate limits will automatically update.”
Guarantee: “Specified rate limits are not guaranteed and actual capacity may vary.”
When a limit is hit: “If you hit a spend-based rate limit, the API returns a 429 RESOURCE_EXHAUSTED error.”
Batch price: “Pricing: Batch API usage is priced at 50% of the standard interactive API cost for the equivalent model.”
Batch window: “The target turnaround time is 24 hours, but in majority of cases, it is much quicker.”
How the tier is set: “Your tier is based on cumulative spend on the xAI API since January 1, 2026. Tiers unlock automatically as your spend increases.”
Tier changes: “Once you qualify for a tier, you stay there permanently; tiers never downgrade.”
What is limited: “Every xAI API team has per-model rate limits on two dimensions: requests per second (RPS) and tokens per minute (TPM).”
Requests per second: “Your per-second limit is derived from your per-minute request budget (RPM / 48, with a minimum of 2)”
What counts: “Cached prompt tokens (still count toward TPM, though they are billed at a reduced rate)”
When a limit is hit: “Exceeding any limit returns a 429 Too Many Requests error.”
Published The provider’s documentation: rate limits page, read .
DeepSeek API rate limits
Limits by model and tier
Model
Tier
Parallel requests
deepseek-flash
All tiers
2,500
deepseek-v4-pro
All tiers
500
Who the limit applies to: “Concurrency limits are calculated at the account level, regardless of which API Key is used”
When a limit is hit: “when the concurrency limit is exceeded, you will receive an HTTP 429 error code”
Raising limits: “If you need higher concurrency, you can submit a capacity expansion request . We will match the appropriate concurrency based on your actual business needs. There is no additional cost for capacity expansion.”
Off-peak price: “(2) Off-peak rates are half of the peak rates. Peak hours are 01:00 - 04:00 and 06:00 - 10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other hours are off-peak, including weekends and Chinese public holidays in full.”
How the tier is set: “Free mode lets you create API keys and use included monthly usage within the limits shown on the Limits page.”
How the tier is set: “Pay-as-you-go lets you extend usage beyond included monthly usage, and this extra usage draws from your prepaid credits. It isn't a separate plan for the API.”
What is limited: “Tokens per minute: the token throughput limit for the model.”
Requests per second: “Requests per second: the request throughput limit for the model.”
Where the numbers are: “API rate limits define how much traffic your Organization can send. They are shown in”
Spend cap: “If the Organization reaches its monthly spending limit, API access can be suspended until the next month begins or an admin increases the limit.”
Batch price: “Batching allows you to run asynchronous inference on large inputs in parallel, reducing compute costs while running large workloads at a 50% discount.”
Batch window: “Users can set timeout_hours when creating a job, which specifies the number of hours after which the job should expire. It defaults to 24 hours and should be lower than 7 days.”
How the tier is set: “Need higher rate limits? Upgrade to Developer plan to access higher limits, Batch and Flex processing, and more.”
Plan in the table: “Note that the limits shown below are the base limits for the Developer plan, and higher limits are available for select workloads and enterprise use cases.”
Cache and limits: “Cached tokens do not count towards your rate limits.”
Who the limit applies to: “Rate limits apply at the organization level, not individual users.”
Input and output limits: “In addition to the combined TPM limit, some organizations are also subject to separate per-minute limits on input tokens (ITPM) and output tokens (OTPM).”
When a limit is hit: “When you exceed rate limits, our API returns a 429 Too Many Requests HTTP status code.”
Batch window: “We allow for setting a batch window from 24 hours to 7 days and recommend setting a longer batch window allow us more time to complete your batch jobs instead of expiring them.”
Batch and cache: “Note: The batch discount does not stack with prompt caching discounts. All batch tokens are billed at the 50% batch rate regardless of cache status.”
Batch and limits: “with 50% lower cost, no impact to your standard rate limits, and 24-hour to 7 day processing window.”
How the tier is set: “New accounts receive $5 in free credits after adding a verified payment method. These credits expire 30 days after they're granted and can be used across all public models.”
How the tier is set: “Purchase credits (Pay as You Go) from the Billing tab in the Cloud Console. Your first purchase moves you to the Developer tier with significantly higher rate limits and no hourly or daily token caps.”
Daily limits: “Hourly and daily restrictions don't apply to Developer (Pay as You Go) tier users. You can use as many tokens as needed within your budget.”
Two limits: “Cerebras uses a dual-bucket model: every organization has both an uncached token limit and a total token limit.”
Cache and limits: “Cached tokens don't count toward your uncached TPM limit, so a higher cache hit rate lets you process far more total tokens within the same uncached limit.”
How limits refill: “We use the token bucketing algorithm for rate limiting, which means your capacity replenishes continuously rather than resetting at fixed intervals.”
Estimate before the request: “If this estimated token consumption would exceed your available token quota, the request is rate limited before processing begins.”
When a limit is hit: “If you exceed your rate limits, you will receive a 429 Too Many Requests error. The error message will indicate whether your uncached or total token limit was exceeded.”
Published The provider’s documentation: rate limits page, read .
Numbers are the providers’ published limits, not our measurements of what an account gets. Where a provider shows its numbers only inside the account console, the quote says so and the table is absent. The task cost calculator checks an average load against these tiers; the rate limit planner finds the tier that finishes a batch job before a deadline.