Xiaomi, China · Released September 21, 2026New
MiMo-V2.6 Flash has 310.8B parameters (15B active per token). In BF16 its weights alone take 577 GiB; quantized to 4-bit, about 174 GiB. Each 32K-token request adds 0.73 GiB of KV cache. Cheapest way to run it today: 8 × GeForce RTX 3090 (24 GB) at $1.14 per hour.
Size your own setup in the calculator →Weights plus the KV cache of one request (BF16 cache), in GiB. Add 1 to 3 GiB for the inference engine itself.
| Weights format | Weights only | + 8K context | + 32K context | + 128K context | + 1M context |
|---|---|---|---|---|---|
| MXFP4 experts + FP8 (native) | 160 | 160 | 161 | 163 | 182 |
| BF16 / FP16 | 577 | 577 | 578 | 580 | 600 |
| FP8 / INT8 | 290 | 290 | 290 | 292 | 312 |
| 4-bit (GGUF Q4_K_M) | 174 | 174 | 174 | 177 | 196 |
Exact size of each quantization in ggml-org/MiMo-V2.6-Flash-RL-GGUF (opens in a new tab), and the smallest GPU that runs it with llama.cpp, Ollama or LM Studio and an 8K-token context.
| Quantization | File size | Smallest GPU |
|---|---|---|
| MXFP4 | 156 GiB | More than 96 GB: see the calculator |
| Q2_K | 118 GiB | More than 96 GB: see the calculator |
Read every day from Hugging Face. Split files are added together; vision projectors are left out.
Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.
| GPU | MXFP4 experts + FP8 (native) | BF16 / FP16 | FP8 / INT8 | 4-bit (GGUF Q4_K_M) |
|---|---|---|---|---|
| GeForce RTX 4090 24 GB | 8 | 32 (several servers) | 16 (several servers) | 16 (several servers) |
| GeForce RTX 5090 32 GB | 6 | 24 (several servers) | 16 (several servers) | 6 |
| RTX PRO 6000 Blackwell 96 GB | 2 | 7 | 4 | 2 |
| Mac M5 Max 128 GB 96 GB usable | 2 | 7 | 4 | 2 |
| NVIDIA H100 80 GB | 3 | 16 (several servers) | 5 | 3 |
| NVIDIA H200 141 GB | 2 | 5 | 3 | 2 |
| NVIDIA B200 (HGX) 180 GB | 2 | 4 | 2 | 2 |
For each GPU: the fewest cards that fit MiMo-V2.6 Flash with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-10-10).
On-demand prices from RunPod, Vast.ai, Verda and Azure, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.
Architecture: MoE 256 experts · 9 global-attention layers (4 KV heads) + 39 sliding-window layers (128 tokens, 8 KV heads), keys 192, values 128. Its KV cache is 10.3× smaller than if every layer used full attention, because most layers only keep a short window of recent tokens. Only 15B parameters are active per token, but all 311B must sit in memory: VRAM depends on the total, speed on the active part.
| Context per request | BF16 cache | FP8 cache |
|---|---|---|
| 8K tokens | 0.20 GiB | 0.10 GiB |
| 32K tokens | 0.73 GiB | 0.36 GiB |
| 128K tokens | 2.84 GiB | 1.42 GiB |
| 1M tokens | 22.5 GiB | 11.3 GiB |
In BF16 the weights alone take 577 GiB (620 GB). In FP8 that is 290 GiB, and about 174 GiB with 4-bit quantization (Q4_K_M); the official MXFP4 experts + FP8 checkpoint is 160 GiB. Each request then adds KV cache: 0.20 GiB at 8K tokens and 0.73 GiB at 32K (BF16 cache).
No. Even in 4-bit it needs 16 (several servers) RTX 4090s. In 4-bit it needs 3 H100 GPUs.
On 2026-10-10, the cheapest on-demand setup is 8 × GeForce RTX 3090 (24 GB) with MXFP4 experts + FP8 weights on Vast.ai, at $1.14 per hour (about $835.12 per month). Next: 4 × RTX 6000 Ada (48 GB) with MXFP4 experts + FP8 weights on Vast.ai, at $1.98 per hour. Sized for one 8K-token request; prices are checked every hour.
With one request and an 8K-token context: 16 (several servers) in BF16, 5 in FP8 and 3 in 4-bit (vLLM, 90% of memory usable). Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.
For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/mimo-v2-6-flash.json.
Xiaomi, China · Released September 21, 2026[](https://studiotvai.com/vram-requirements/mimo-v2-6-flash)Same family (large moe): MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · MiMo-V2.6 Pro · Kimi K3 · Mistral Large 4 · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Atria Dawn Preview · Intern S2 397B
All models: Llama 3.1 8B · Mistral Small 3.2 24B · Qwen3 32B · Llama 3.3 70B · Llama 3.1 405B · Qwen3.5 4B · Qwen3.5 9B · Qwen3.8 27B · Qwen3.6 35B-A3B · Qwen3.5 122B-A10B · Qwen3.8-Flash-Next · LFM2.5 8B-A1B · LFM2 24B-A2B · Gemma 4 12B · Gemma 4 31B · Gemma 4 26B-A4B · gpt-oss 20B · gpt-oss 120B · Llama 4 Scout 17B-16E · Llama 4 Maverick 17B-128E · Kolibri-1 · Nemotron 3 Nano 30B-A3B · Nemotron 3 Super 120B-A12B · MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · MiMo-V2.6 Pro · Kimi K3 · Mistral Large 4 · Qwen3-Coder-Next 80B-A3B · Qwen3-Coder 30B-A3B · Qwen3-Coder 480B-A35B · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Ornith 1.5 35B-A3B · Ornith 1.5 9B · Atria Dawn Preview · Intern S2 397B · Clef · Clef Flash · D1 3B · Humanizer · JEV 27B VL · Mellum2.1 12B-A2.5B · LightOnOCR-3 4B · Spark-X2.5 4B · Agnes-3.0-Qwen
Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.