deepseek-v3.1:671b-cloud

Upstream reference: DeepSeek-V3.1

Publisher: DeepSeek

The publisher documents a mixture-of-experts model with thinking and non-thinking modes selected through the chat template, plus version-specific tool-call formats.

Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.

Published upstream specifications

Total parameters
671 billion Source
Active parameters
37 billion Source
Published context
128K tokens, publisher notation Source
Upstream license label
MIT Source

Publisher-described uses

  • Text reasoning, code and tool-oriented applications. Source

Reference limitations

  • Thinking mode and tool-call behavior depend on the template and surrounding application; a cloud alias does not establish them. Source

A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.

Publisher sources

Links

Recorded references do not establish downloadable weights, licensing or current availability.

671.0B
parameters
FP8
quantization
deepseek2
architecture
163840
context
mit
license
2025-08-21
released

Where deepseek-v3.1:671b-cloud comes from

Company / lab

DeepSeek

See every model and family from this maker.

Model family

deepseek-v3.1

Other versions and sizes in this line.

completionthinkingtools

Named in 2 observation pages

Browse the Observatory

AI-generated & AI-scored. Every observation on this site is produced and judged by AI, and AI makes mistakes. Treat each result as a recorded observation under a declared protocol, as of its date. Method and marker legend