Model library / XiaomiMiMo
MiMo-V2.6-Flash-RL is a 159.4 billion-parameter sparse MoE causal model for multimodal long-context reinforcement-learning 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-RL is a 159.4 billion-parameter causal model using a sparse mixture-of-experts architecture. XiaomiMiMo describes it as an efficiency-balanced checkpoint aimed at scaling reinforcement learning for self-improvement, supporting text, image, video and audio inputs and a one-million token context window, and enables long-horizon reasoning across multi-session agent runs.
In deployment the model occupies roughly 160 billion parameters, stored in fp8 format, and routes each token through 8 of 256 experts. It accepts sequences up to 1,048,576 tokens, making it suitable for very long documents or tool traces. The code and weights are released under the MIT licence.
Specification
| Repository | XiaomiMiMo/MiMo-V2.6-Flash-RL |
| Publisher | XiaomiMiMo |
| Published | 21 September 2026 |
| Architecture | MiMoV2ForCausalLM |
| Model type | mimo_v2 |
| Parameters | 159,358,725,504 |
| 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 | 296.8 GiB | Full precision as released |
| fp8 | 148.4 GiB | 8-bit, near-lossless on most models |
| int4 | 74.2 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 | 8 | 66.7 GiB | about 485,401 tokens |
| RTX 5090 32GB | 4 | 19.1 GiB | about 138,788 tokens |
| L40S 48GB | 4 | 76.7 GiB | about 558,218 tokens |
| A100 80GB | 2 | 52.9 GiB | about 384,912 tokens |
| H100 80GB | 2 | 52.9 GiB | about 384,912 tokens |
| RTX PRO 6000 96GB | 2 | 81.7 GiB | about 594,627 tokens |
| H200 141GB | 1 | 38.3 GiB | about 278,598 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-RL \ --max-model-len 483328 \ --gpu-memory-utilization 0.90
Questions
The weights take about 86 GiB at the released precision. That fits on one H200 141GB, which leaves roughly 38 GiB for the key-value cache.
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-RL is a 524-billion-parameter sparse MoE model designed for reinforcement-learning-driven self-improvement across t…
AliceAI-Foundation-80B-A3B-Base is an 81-billion-parameter autoregressive language model with mixture-of-experts layers, trained …
Ling-3.0-flash-Fin is a 127.5 billion-parameter finance-enhanced language model for long-context financial research and analysis.
Motif 3 is a 314.8 billion-parameter decoder-only mixture-of-experts language model with 262 k token context, released under MIT.
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.