Evidence100%Authoritative
FactConfirmedProduct·November 6, 2025
Cerebras Delivers 3,000 Tokens/Second Inference for OpenAI's gpt-oss-120B Open-Weight Model
Cerebras announced support for OpenAI's first open-weight reasoning model (gpt-oss-120B) on its inference service, running at ~3,000 tokens per second — claimed to be ~55x faster and ~60x cheaper than Anthropic's Claude 4 Opus.
Evidence Strength
Evidence100%Authoritative
Backed by official company doc
Reported by 2 independent publishers
Includes official or primary source
Key Development
High-significance development (rated 8/10)
Covered by 4 sources
Confirmed — verified event
Insights
First tracked
August 5, 2025
Last updated
November 6, 2025
Sources
4 sources
Related Developments
OpenAI Signs $10B+ Multiyear Compute Deal with CerebrasOpenAI GPT-5.3-Codex-Spark Powered by Cerebras Launches in Research PreviewGPT-5.3-Codex-Spark Ships as First Model Running on Cerebras WSE-3 for OpenAIOklahoma City AI Datacenter Ribbon-Cutting with 44+ ExaflopsCS-3 vs. NVIDIA DGX B200 Blackwell Benchmarks Published
Sources (4)
Source Timeline
OpenAI GPT-OSS 120B Benchmarked – NVIDIA Blackwell vs. CerebrasCerebras·Nov 6, 2025- OpenAI GPT OSS 120B Runs Fastest on CerebrasCerebras·Aug 6, 2025
Cerebras delivers blazing speed for OpenAI’s new open-model with 3,000 tokens per secondSiliconANGLE·Aug 5, 2025- Cerebras Launches OpenAI’s gpt-oss-120B at a Blistering 3,000 tokens/secCerebras·Aug 5, 2025
Evidence Strength
Evidence100%Authoritative
Backed by official company doc
Reported by 2 independent publishers
Includes official or primary source
Key Development
High-significance development (rated 8/10)
Covered by 4 sources
Confirmed — verified event
Insights
First tracked
August 5, 2025
Last updated
November 6, 2025
Sources
4 sources
Related Developments
OpenAI Signs $10B+ Multiyear Compute Deal with CerebrasOpenAI GPT-5.3-Codex-Spark Powered by Cerebras Launches in Research PreviewGPT-5.3-Codex-Spark Ships as First Model Running on Cerebras WSE-3 for OpenAIOklahoma City AI Datacenter Ribbon-Cutting with 44+ ExaflopsCS-3 vs. NVIDIA DGX B200 Blackwell Benchmarks Published