Cerebras Inference adds dual-bucket rate limits for uncached and total tokens

Cerebras now uses a dual-bucket rate limiting model. Every organization has two independent token limits:

• Uncached TPM — tokens that require full compute (cache misses).
• Total TPM — uncached + cached tokens combined.

Cached tokens don't count toward your uncached limit, so a higher cache hit rate lets you process far more total tokens within the same uncached budget. 429 errors now indicate which bucket was exceeded. See Uncached vs. Total Tokens for details.

This update is rolling out gradually and will be fully in effect for all organizations on August 17, 2026. During the transition, the uncached limit may not yet appear in your console — this is expected.