Model library / XiaomiMiMo
MiMo-V2.6-Flash-MOPD is a 310.8 B parameter sparse-MoE model that merges domain-specialised teachers to reduce tool-call repetition in long-horizon tasks.
8 of 256 experts active per token
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
MiMo-V2.6-Flash-MOPD is the MOPD2 upgrade of the MiMo-V2.6-Flash-RL checkpoint. XiaomiMiMo describes it as a distillation of mixRL and SFT teachers, extending capability to domains where reliable training-time verification is difficult, such as long-horizon game development, scientific research and embodied intelligence. It also addresses tool-call repetition in agentic settings.
In practice the model contains 310.8 billion parameters organised as a 256-expert mixture of experts with eight experts active per token, yielding about 15 billion activated parameters. It supports a 1,048,576-token context window, is released under the MIT licence, and uses fp8 weight format. Deployment should consider its size, sparse execution, and licence terms.
Specification
| Repository | XiaomiMiMo/MiMo-V2.6-Flash-MOPD |
| Publisher | XiaomiMiMo |
| Published | 27 September 2026 |
| Architecture | MiMoV2ForCausalLM |
| Model type | mimo_v2 |
| Parameters | 310,756,322,688 |
| Experts | 256 total, 8 active per token |
| Layers | 48 |
| Hidden size | 4,096 |
| Attention heads | 64, 4 key/value heads |
| Context length | 1,048,576 tokens |
| Vocabulary | 152,576 tokens |
| Quantisation | fp8 |
| Licence | mit |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 578.8 GiB | Full precision as released |
| fp8 | 289.4 GiB | 8-bit, near-lossless on most models |
| int4 | 144.7 GiB | 4-bit, smallest footprint |
The cache costs 144 KiB per token at 16-bit, so the full 1,048,576-token context of one request needs about 144.0 GiB. Concurrency multiplies that number, not the weights.
That figure is an upper bound: this model uses sliding-window attention with a 128-token window on some layers, so long requests cache less than this.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | more than 8 | — | — |
| RTX 5090 32GB | 8 | 49.4 GiB | about 359,380 tokens |
| L40S 48GB | 4 | 1.8 GiB | about 12,768 tokens |
| A100 80GB | 4 | 117.0 GiB | about 851,629 tokens |
| H100 80GB | 4 | 117.0 GiB | about 851,629 tokens |
| RTX PRO 6000 96GB | 2 | 6.8 GiB | about 49,177 tokens |
| H200 141GB | 2 | 87.8 GiB | about 639,001 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve XiaomiMiMo/MiMo-V2.6-Flash-MOPD \ --tensor-parallel-size 8 \ --max-model-len 356352 \ --gpu-memory-utilization 0.90
Questions
The weights take about 161 GiB at the released precision. No single card in the table above holds that, so it needs several GPUs or a lower precision.
The repository declares mit. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 1,048,576 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
MiMo-V2.6-Pro-MOPD is a 1.02 trillion-parameter sparse Mixture-of-Experts language model with multimodal support, released by Xia…
MiMo-V2.6-Flash-RL is a 159.4 billion-parameter sparse MoE causal model for multimodal long-context reinforcement-learning tasks.
MiMo-V2.6-Pro-RL is a 524-billion-parameter sparse MoE model designed for reinforcement-learning-driven self-improvement across t…
Ling-3.1-flash is a 572.1 billion-parameter hybrid reasoning language model with a 262,144-token context window.
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.