MiMo · Over 80B · Mixture of experts
MiMo V2 Flash
309.79 billion parameters in total, but only 16.05 billion are read for each token. That split is the whole point of the design: it costs the memory of a large model and the speed of a small one. Grouped-query attention keeps the cache small, so long contexts cost less here than on models of the same size.
Memory needed, by quantisation
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantisation | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 306.55 GB | 0.02 GB | 307.42 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 236.58 GB | 0.02 GB | 237.45 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 205.21 GB | 0.02 GB | 206.07 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 174.19 GB | 0.02 GB | 175.06 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 153.27 GB | 0.02 GB | 154.14 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 141.01 GB | 0.02 GB | 141.88 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 110.36 GB | 0.02 GB | 111.23 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. This model uses sliding-window attention, so most layers stop growing past 128 tokens and the bill flattens out.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 0.02 GB | 174.93 GB | no | no | no |
| 8k | 0.02 GB | 175.06 GB | no | no | no |
| 16k | 0.02 GB | 175.31 GB | no | no | no |
| 32k | 0.02 GB | 175.81 GB | no | no | no |
| 64k | 0.02 GB | 176.81 GB | no | no | no |
| 128k | 0.02 GB | 178.81 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs MiMo V2 Flash
3 of 118 consumer devices run it at a quantisation worth using, at an 8k context window. Another 7 can load it only by compressing the weights far enough to damage the model, marked with a warning below.
| Device | Memory | Verdict | Quantisation | Used | Speed |
|---|---|---|---|---|---|
| Apple M3 Ultra 512GB Apple | 512 GB | Runs great | Q8_0 | 307.12 GB | 40.2 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs great | Q4_K_M | 174.76 GB | 70.6 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs great | Q3_K_M | 141.58 GB | 85.2 t/s |
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | IQ2_XXS ! | 75.16 GB | 380 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs great | IQ2_XXS ! | 74.86 GB | 161.4 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs great | IQ2_XXS ! | 74.86 GB | 161.4 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs great | IQ2_XXS ! | 74.86 GB | 110.1 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Runs great | IQ2_XXS ! | 74.86 GB | 80.7 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs great | IQ2_XXS ! | 75.16 GB | 57.9 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Runs great | IQ2_XXS ! | 75.16 GB | 43 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Partial offload 12/48 | — | 175.06 GB | 4.6 t/s |
| RTX A6000 NVIDIA | 48 GB | Partial offload 12/48 | — | 175.06 GB | 4.6 t/s |
| Radeon PRO W7900 AMD | 48 GB | Partial offload 12/48 | — | 175.06 GB | 4.6 t/s |
| GeForce RTX 5090 NVIDIA | 32 GB | Partial offload 8/48 | — | 175.06 GB | 4.2 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Partial offload 6/48 | — | 175.06 GB | 4 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Partial offload 6/48 | — | 175.06 GB | 4 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Partial offload 6/48 | — | 175.06 GB | 4 t/s |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Partial offload 6/48 | — | 175.06 GB | 4 t/s |
| RTX A5000 NVIDIA | 24 GB | Partial offload 6/48 | — | 175.06 GB | 4 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Partial offload 6/48 | — | 175.06 GB | 4 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Partial offload 5/48 | — | 175.06 GB | 3.9 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 5060 Ti 16GB NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 4080 NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 4070 Ti SUPER NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 4060 Ti 16GB NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 4090 Laptop NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| RTX A4000 NVIDIA | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Radeon RX 9070 XT AMD | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Radeon RX 9070 AMD | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Radeon RX 7900 GRE AMD | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Radeon RX 7800 XT AMD | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Radeon RX 7600 XT AMD | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Radeon RX 6900 XT AMD | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Radeon RX 6800 AMD | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| Arc A770 16GB Intel | 16 GB | Partial offload 3/48 | — | 175.06 GB | 3.7 t/s |
| GeForce RTX 5070 NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 5060 Ti 8GB NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 5060 NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 4070 Ti NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 4070 SUPER NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 4070 NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 4060 Ti 8GB NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 4060 NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3080 Ti NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3080 12GB NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3080 10GB NVIDIA | 10 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3070 Ti NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3070 NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3060 Ti NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3060 12GB NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3050 8GB NVIDIA | 8 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 3050 6GB NVIDIA | 6 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 2080 Ti NVIDIA | 11 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce RTX 2060 12GB NVIDIA | 12 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce GTX 1080 Ti NVIDIA | 11 GB | Partial offload 2/48 | — | 175.06 GB | 3.6 t/s |
| GeForce GTX 1660 SUPER NVIDIA | 6 GB | Partial offload 1/48 | — | 175.06 GB | 3.6 t/s |
What to buy to run MiMo V2 Flash
The cheapest hardware that runs it at a quantisation worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
Apple M3 Ultra 256GB
$5,599 whole machine
Q4_K_M · uses 174.76 GB of its 256 GB · about 70.6 tokens/s
Best value
Apple M3 Ultra 512GB
$9,499 whole machine
Q8_0 · about 40.2 tokens/s · 819 GB/s
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| Apple M3 Ultra 256GB Apple · whole machine | $5,599 | 256 GB | Q4_K_M | 70.6 t/s | Check price |
| Apple M3 Ultra 512GB Apple · whole machine | $9,499 | 512 GB | Q8_0 | 40.2 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
Where it sits in the catalogue
Source: XiaomiMiMo/MiMo-V2-Flash on Hugging Face. Downloaded 88k times in the last month. Published 2025-12-16. Architecture figures are read from the repository's own configuration file, so they move when the model does.