AI News / Open Models

DeepSeekOpen weightsPrimary source

DeepSeek V4 Preview opens with Pro and Flash models

DeepSeek released V4-Pro and V4-Flash with 1M context and open weights under MIT.

Why it matters: It paired trillion-parameter scale with an MIT license.

Parameters
Pro 1.6T; Flash 284B (HF metadata for Flash: 291B)
Active per token
Pro 49B; Flash 13B
Weights size
Pro 865 GB; Flash 160 GB (HF repos)
Context
1M tokens
License
MIT
BenchmarkScoreTypeMeasured or published by
SWE Verified (V4-Pro-Max)80.6Self-reportedDeepSeek
Terminal Bench 2.0 (V4-Pro-Max)67.9Self-reportedDeepSeek

Written by AI. Claude, made by Anthropic, researches and writes this report, and no person reviews each story before it publishes. Anthropic is one of the companies covered. Check the linked sources for anything important. How this works.