Chutes · Rate Limits

Chutes Rate Limits

Chutes enforces per-account quotas and concurrency limits on public inference through llm.chutes.ai/v1, governed by available subnet capacity, the account's plan (pay-as-you-go vs. Plus/Pro monthly budget), and per-model availability. Free / subsidized models carry the tightest throttles. Private (dedicated) chutes are limited by the GPU capacity provisioned for the deployment rather than a shared quota. Specific per-model RPM/TPM values are not reconciled in this artifact.

Chutes Rate Limits is the machine-readable rate-limit profile for Chutes on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 6 rate-limit definitions, measuring requests, tokens, concurrent_requests, spend, and gpu_capacity.

The profile also includes 3 backoff/retry policies defined and response codes documented for throttled.

Tagged areas include AI, LLM, Inference, Serverless, and GPU.

6 Limits Throttle: 429
AILLMInferenceServerlessGPUBittensorRate LimitingQuotasThrottling

Limits

Requests Per Minute (RPM) account
requests
see provider documentation
Per-model and per-plan; varies with subnet capacity.
Tokens Per Minute (TPM) account
tokens
see provider documentation
Per-model and per-plan; varies with subnet capacity.
Concurrency account
concurrent_requests
see provider documentation
Concurrent in-flight inference requests per account.
Free / Subsidized Model Quota account
requests
see provider documentation
Free and subsidized models carry tighter throttles than paid models.
Monthly Budget Quota account
spend
plan dependent
Plus/Pro monthly plans include a fixed budget; usage beyond it is discounted per-token.
Private Chute Capacity deployment
gpu_capacity
provisioned per deployment
Dedicated chutes are bounded by the GPU capacity provisioned, not a shared quota.

Policies

Plan-Tiered Limits
Quotas and budgets raise as accounts move from pay-as-you-go to Plus, Pro, and Enterprise plans.
Subnet Capacity
Public inference throughput depends on live Bittensor Subnet 64 miner capacity for the requested model.
Backoff Strategy
Clients should implement exponential backoff with jitter and honor Retry-After on 429 responses.

Sources