MiMo V2 Flash needs about 203.7 GB of memory at Q4_K_M (a 187.1 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 27 quants, from 0.5 GB to 328.2 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 64.4 GB.
MiMo V2 Flash – a 309B parameter Mixture‑of‑Experts LLM
MiMo V2 Flash is a 309B parameter Mixture‑of‑Experts language model with 15B active parameters. It supports up to 262144 context tokens and is designed for reasoning and tool‑using workflows. The model is released under the MIT license.
Good at
- Long‑context reasoning
- Tool‑based task execution
- Fast inference with MTP
Limits
- Requires large VRAM for higher‑quality quantizations
- No vision support
- Context limited to 262144 tokens
Best for
- Agentic workflow automation
- Complex reasoning tasks
- Large‑scale inference on high‑VRAM GPUs
Who it’s for. Developers building agentic applications or research teams needing a high‑capacity MoE model. The model is open‑source and can be run with llama.cpp or other GGUF‑compatible engines.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.5 GB | 17.1 GB | Too big | Get file → | |
| IQ1_S | 63.3 GB (2 parts) | 79.9 GB | Too big | Get file → | |
| IQ1_M | 65.8 GB (2 parts) | 82.4 GB | Too big | Get file → | |
| IQ2_XXS | 73.5 GB (2 parts) | 90.1 GB | Too big | Get file → | |
| IQ2_XS | 84.9 GB (3 parts) | 101.5 GB | Too big | Get file → | |
| IQ2_S | 85.2 GB (3 parts) | 101.8 GB | Too big | Get file → | |
| IQ2_M | 96.7 GB (3 parts) | 113.3 GB | Too big | Get file → | |
| Q2_K | 107.7 GB (3 parts) | 124.3 GB | Too big | Get file → | |
| Q2_K_L | 108.3 GB (3 parts) | 124.9 GB | Too big | Get file → | |
| IQ3_XXS | 121.5 GB (4 parts) | 138.1 GB | Too big | Get file → | |
| IQ3_XS | 126.4 GB (4 parts) | 143.0 GB | Too big | Get file → | |
| Q3_K_S | 133.7 GB (4 parts) | 150.3 GB | Too big | Get file → | |
| IQ3_M | 140.3 GB (4 parts) | 156.9 GB | Too big | Get file → | |
| Q3_K_M | 140.3 GB (4 parts) | 156.9 GB | Too big | Get file → | |
| Q3_K_L | 146.1 GB (4 parts) | 162.7 GB | Too big | Get file → | |
| Q3_K_XL | 146.6 GB (4 parts) | 163.2 GB | Too big | Get file → | |
| IQ4_XS | 164.8 GB (5 parts) | 181.4 GB | Too big | Get file → | |
| IQ4_NL | 174.4 GB (5 parts) | 191.0 GB | Too big | Get file → | |
| Q4_0 | 177.3 GB (5 parts) | 193.9 GB | Too big | Get file → | |
| Q4_K_S | 180.5 GB (5 parts) | 197.2 GB | Too big | Get file → | |
| Q4_K_M | 187.1 GB (5 parts) | 203.7 GB | Too big | Get file → | |
| Q4_K_L | 187.6 GB (5 parts) | 204.2 GB | Too big | Get file → | |
| Q4_1 | 193.4 GB (5 parts) | 210.0 GB | Too big | Get file → | |
| Q5_K_S | 212.6 GB (6 parts) | 229.2 GB | Too big | Get file → | |
| Q5_K_M | 219.2 GB (6 parts) | 235.8 GB | Too big | Get file → | |
| Q6_K | 253.5 GB (7 parts) | 270.1 GB | Too big | Get file → | |
| Q8_0 | 328.2 GB (9 parts) | 344.8 GB | Too big | Get file → |
For 8 GB VRAM, use the 0.5 GB unknown quant. For 16 GB VRAM, IQ1_S (63.25 GB) or IQ1_M (65.83 GB) are suitable. For 24 GB VRAM, Q2_K (107.66 GB) or Q3_K_S (133.69 GB) provide a balance of quality and memory usage. Estimates are approximate.
What GPU do you need for MiMo V2 Flash?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 17.1 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 17.1 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 17.1 GB in use).
Which MiMo V2 Flash quant should you choose?
Q4_K_M is the usual starting point for MiMo V2 Flash: it balances file size and quality well. Here it is a 187.1 GB file.
The Q8_0 file is 328.2 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is the first one in the table at 0.5 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about MiMo V2 Flash
How much VRAM does MiMo V2 Flash need?
About 203.7 GB at Q4_K_M with an 8K context window. That is the 187.1 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run MiMo V2 Flash on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 17.1 GB, so you would need a larger GPU or CPU offloading.
What is the file size of MiMo V2 Flash GGUF?
It depends on the quant: file is 0.5 GB, IQ1_S is 63.3 GB, IQ1_M is 65.8 GB, IQ2_XXS is 73.5 GB, IQ2_XS is 84.9 GB, IQ2_S is 85.2 GB and 21 more.
What context length does MiMo V2 Flash support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 16.1 GB at 8K and 64.4 GB at 32K.
Where can I download MiMo V2 Flash GGUF?
The files are on Hugging Face in the bartowski/XiaomiMiMo_MiMo-V2-Flash-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does MiMo V2 Flash use?
Hugging Face lists the license as mit. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run MiMo V2 Flash on 8 GB VRAM?
Yes, the 0.5 GB unknown quant can be loaded on 8 GB VRAM, but performance may be limited.
Does MiMo V2 Flash support vision tasks?
No, the model does not include vision capabilities.