Announcement
Cerebellum Labs announces custom-model hosting without the GPU-hour bill
The company says Cerebellum Cloud bills per token, with no idle charges and $25 in starter credits.
TLDR
According to Cerebellum Labs, its new Cloud service serves open-weight text models up to 70 billion parameters through OpenAI-compatible endpoints, charging for input and output tokens. Users can import weights from Hugging Face, S3 or GCS. The company says idle time costs nothing, but there’s a wake-up wait: the first request after inactivity takes longer while the model starts.
Combined views
18K
3 Sources, first seen ago

Comments