Alex Getmanearly AI signals

MiniMax M2.5

MiniMax

Released Updated
Context window
205K
204,800 tokens
Input / Output
text → text

Timeline

  1. · Weights version · evidence link
  1. · Evidence links · GitHub search · MiniMax · commits · ThinkInAIXYZ/PublicProviderConf · @zerob13
    Excerpt
    updated: Qwen3.5 9B, Gemma 4 26B A4B IT, MiniMax-M2.5 - LLM Gateway: +7 models, 8 updates
  2. · Evidence links · GitHub search · MiniMax · pulls · lostcoords/yuke · @yuke-catalog-action[bot]
  3. · Evidence links · GitHub search · MiniMax · pulls · inference-sim/blis-vllm · @sriumcp
    Excerpt
    and 289 us over NVLink for 8192 tokens (MiniMax-M2.5, ib-400g), and the assumption is disclosed either way. Fails if the rack is always the hardware's
  4. · Evidence links · GitHub search · MiniMax · pulls · inference-sim/accuracy-dashboard · @sriumcp
    Excerpt
    catalog model name, and the NVFP4 points use `minimax-m2.5-nvfp4` while the entries say `minimax-m2.5`. Should the reader match on the model family, or should the
  5. · Evidence links · GitHub search · MiniMax · pulls · AMD-AGI/AKLR · @zhenhuang12
    Excerpt
    glm-5.3 53, deepseek-v4-flash 45, deepseek-r1 40, deepseek-v3 40, minimax-m2.5 32, muse-spark 11, minimax-m3.1 5. 17 scenarios name two models.
Specifications

MiniMax M2.5 is a model from MiniMax with open weights, released on 2026-02. The context window is 204,800 tokens.

Context window
204,800
Input
text
Output
text
Weights
Open

Questions

What is the context window of MiniMax M2.5?

The context window is 204,800 tokens.

Are the weights of MiniMax M2.5 publicly available?

Yes. MiniMax publishes the weights.

When was MiniMax M2.5 released?

On 2026-02.