Xiaomi, released 2025-12-16. Sparse MoE decoder, 309B total, 15B active, 256K context. Attention: 5:1 sliding-window/global attention.
Sparse omnimodal MoEXiaomi 🇨🇳 · April 22, 2026
Xiaomi MiMo-V2.5 (310B)
Omnimodal sparse MoE model that extends the MiMo-V2-Flash backbone with vision and audio encoders.
Key detail
Adds native image, video, and audio encoders to the MiMo-V2-Flash-style sparse MoE backbone with MTP support.Total params
310B
Active / token
15B
Context
1M
Fact sheet
- Scale
- 310B total, 15B active (4.8% active)
- Context window
- 1,048,576 tokens
- Decoder type
- Sparse omnimodal MoE
- Attention
- 5:1 sliding-window/global attention
- Layer mix
- 39 sliding-window + 9 global
- KV cache / token (bf16)
- 144 KiB · Moderate
- License
- MIT License
- Released
Efficiency
Parameters active per token
4.8%
15B of 310B weights run for each token.
AA Intelligence Index v4.3.2
Source21.8
- General
- 30.4
- Scientific
- 15.5
- Coding
- 21.9
- Agents
- 17.7
Architecture concepts
Sources & resources
Use the ← and → keys to step through releases.
Similar architectures
Xiaomi, released 2026-09-21. Sparse MoE decoder, 309B total, 15B active, 1M context. Attention: GQA with sliding-window/global attention.
Motif Technologies, released 2026-07-21. Sparse MoE decoder, 314B total, 13B active, 256K context. Attention: GDLA with MLA-style latent projections, RoPE, and 3:1 sliding-window/full attention.
MiniMax, released 2026-02-12. Sparse MoE decoder, 230B total, 10B active, 192K context. Attention: GQA with QK-Norm.