Cerebras adds Qwen 3.8 27B to public API endpoints with up to 1,500 tokens per second throughput
EDITOR BRIEF
Cerebras now lists Qwen 3.8 27B as available on its public endpoints for free trial and pay-as-you-go users, with 64K context on free tiers and 128K on paid tiers. The model is served at roughly 1,500 tokens per second, while GPT OSS 120B remains available at around 3,000 tokens per second.
INSIGHTS
The launch highlights growing competition around high-throughput inference as model providers emphasize speed, context length, and API accessibility. Cerebras’ transparency around unpruned models and selective storage quantization may appeal to developers concerned about model fidelity in hosted inference services.
COMMENTS
Discussion
> geekhaus:~$ next read?
Next read recommendations
phreak.ing
Phreak.ing publishes dial codes for an internet phone switch with calls, voicemail, intercom, paging, chat, and radio features
The Verge
The Social Reckoning is flopping at the box office
purplesyringa.moe