'%3e%3crect%20x='16'%20y='51'%20width='68'%20height='27'%20rx='13.5'/%3e%3ccircle%20cx='32'%20cy='53'%20r='15'/%3e%3ccircle%20cx='53'%20cy='43'%20r='20'/%3e%3ccircle%20cx='71'%20cy='57'%20r='12'/%3e%3c/g%3e%3c/svg%3e){kind=small}
Cloud Changelog
Cerebras Inference changelog
62 announcements since February 2025
Free. Unsubscribe any time.
RSS feed for Cerebras Inference
November 20261
October 20262
September 20263
August 20262
July 20268
zai-glm-4.7 deprecated August 17, 2026, keeps default logprobs shape until then
Image inputs limited to 15,000 pixels per side, larger images return 413
Predicted Outputs available by request on dedicated endpoints, gpt-oss-120b in preview
gemma-4-31b raises per-request image limit from 5 to 10 on public endpoints
API version 2 is now the default, with stricter validation and reasoning_logprobs
Cerebras Inference adds dual-bucket rate limits for uncached and total tokens
New accounts get $5 in free credits after adding a verified payment method
gemma-4-31b added to Cline, Kilo Code, opencode and VS Code guides
June 20262
May 20261
April 20264
GLM 5, GLM 5.1 and Kimi K2.6 supported on Cerebras Dedicated Endpoints
Cerebras Inference validation errors now return HTTP 400 instead of 422
Cerebras Inference adds prompt_cache_key parameter to chat completions and completions
Chat and text completions accept msgpack-encoded and gzip-compressed request bodies
March 20267
frequency_penalty, presence_penalty and logit_bias available on all models
Selective 4-bit weight-only quantization rolling out for supported models
Chat Completions API supports the developer message role for gpt-oss-120b
Maximum temperature for Chat Completions raised from 1.5 to 2.0
Logprobs computed before temperature scaling, so they no longer vary with temperature
disable_reasoning parameter deprecated, ends July 21, 2026; use reasoning_effort