cerebras/MiniMax-M2-REAP-139B-A10B needs about 87.0 GB of memory at Q4_K_M (a 84.4 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 26 quants, from 0.3 GB to 147.9 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 8.3 GB.
MiniMax-M2-REAP-139B-A10B – a 139B parameter sparse Mixture‑of‑Experts model compressed with REAP
MiniMax-M2-REAP-139B-A10B is a 139B parameter sparse Mixture‑of‑Experts causal language model. It uses REAP pruning to reduce the full 230B model by 40% while keeping near‑lossless performance. The model supports a 196,608 token context length and is compatible with vanilla vLLM.
Good at
- Code generation
- Agentic coding and tool calling
- Math and reasoning
Limits
- No vision support
- Requires large VRAM for full precision
- License is other
Best for
- Local deployment in resource‑constrained environments
- Academic research on sparse models
- Fine‑tuning with limited GPU memory
Who it’s for. Developers and researchers who need a large‑scale language model that fits into smaller GPU budgets. The model is suitable for local inference and academic experimentation where full 230B is impractical.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.3 GB | 2.9 GB | Fits | Get file → | |
| IQ1_S | 29.0 GB | 31.6 GB | Too big | Get file → | |
| IQ1_M | 30.2 GB | 32.8 GB | Too big | Get file → | |
| IQ2_XXS | 33.6 GB | 36.2 GB | Too big | Get file → | |
| IQ2_XS | 38.7 GB | 41.3 GB | Too big | Get file → | |
| IQ2_S | 38.9 GB | 41.5 GB | Too big | Get file → | |
| IQ2_M | 44.1 GB | 46.7 GB | Too big | Get file → | |
| Q2_K | 48.9 GB | 51.5 GB | Too big | Get file → | |
| Q2_K_L | 49.5 GB | 52.1 GB | Too big | Get file → | |
| IQ3_XXS | 55.1 GB (2 parts) | 57.7 GB | Too big | Get file → | |
| IQ3_XS | 57.3 GB (2 parts) | 59.9 GB | Too big | Get file → | |
| Q3_K_S | 60.6 GB (2 parts) | 63.2 GB | Too big | Get file → | |
| IQ3_M | 63.5 GB (2 parts) | 66.1 GB | Too big | Get file → | |
| Q3_K_M | 63.5 GB (2 parts) | 66.1 GB | Too big | Get file → | |
| Q3_K_L | 65.9 GB (2 parts) | 68.5 GB | Too big | Get file → | |
| Q3_K_XL | 66.5 GB (2 parts) | 69.1 GB | Too big | Get file → | |
| IQ4_XS | 74.5 GB (2 parts) | 77.1 GB | Too big | Get file → | |
| IQ4_NL | 78.8 GB (2 parts) | 81.4 GB | Too big | Get file → | |
| Q4_0 | 80.0 GB (3 parts) | 82.6 GB | Too big | Get file → | |
| Q4_K_S | 81.5 GB (3 parts) | 84.1 GB | Too big | Get file → | |
| Q4_K_M | 84.4 GB (3 parts) | 87.0 GB | Too big | Get file → | |
| Q4_1 | 87.3 GB (3 parts) | 89.9 GB | Too big | Get file → | |
| Q5_K_S | 95.9 GB (3 parts) | 98.5 GB | Too big | Get file → | |
| Q5_K_M | 98.8 GB (3 parts) | 101.4 GB | Too big | Get file → | |
| Q6_K | 114.3 GB (3 parts) | 116.9 GB | Too big | Get file → | |
| Q8_0 | 147.9 GB (4 parts) | 150.5 GB | Too big | Get file → |
For 8 GB VRAM, use the IQ1_S quant (file size 28.97 GB, estimated 31.6 GB memory). For 16 GB VRAM, IQ2_XS (38.72 GB, 41.3 GB memory) or Q2_K (48.9 GB, 51.5 GB memory) are suitable. For 32 GB VRAM, Q3_K_M (63.49 GB, 66.1 GB memory) or Q4_K_M (84.43 GB, 87 GB memory) provide a balance of quality and memory usage.
What GPU do you need for cerebras/MiniMax-M2-REAP-139B-A10B?
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: the best quant that fits is the main file (0.3 GB file, about 2.9 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 2.9 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 2.9 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 2.9 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 2.9 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.3 GB file, about 2.9 GB in use).
- 32 GB VRAM: the best quant that fits is IQ1_S (29.0 GB file, about 31.6 GB in use).
- 48 GB VRAM: the best quant that fits is IQ2_M (44.1 GB file, about 46.7 GB in use).
Which cerebras/MiniMax-M2-REAP-139B-A10B quant should you choose?
Q4_K_M is the usual starting point for cerebras/MiniMax-M2-REAP-139B-A10B: it balances file size and quality well. Here it is a 84.4 GB file.
The Q8_0 file is 147.9 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 0.3 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 cerebras/MiniMax-M2-REAP-139B-A10B
How much VRAM does cerebras/MiniMax-M2-REAP-139B-A10B need?
About 87.0 GB at Q4_K_M with an 8K context window. That is the 84.4 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 cerebras/MiniMax-M2-REAP-139B-A10B on an 8 GB GPU?
Yes, with the smaller files: the 0.3 GB file needs about 2.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of cerebras/MiniMax-M2-REAP-139B-A10B GGUF?
It depends on the quant: file is 0.3 GB, IQ1_S is 29.0 GB, IQ1_M is 30.2 GB, IQ2_XXS is 33.6 GB, IQ2_XS is 38.7 GB, IQ2_S is 38.9 GB and 20 more.
What context length does cerebras/MiniMax-M2-REAP-139B-A10B support?
The GGUF header lists a trained context length of 196,608 tokens (192K). The KV cache grows with the context you use: about 2.1 GB at 8K and 8.3 GB at 32K.
Where can I download cerebras/MiniMax-M2-REAP-139B-A10B GGUF?
The files are on Hugging Face in the bartowski/cerebras_MiniMax-M2-REAP-139B-A10B-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 cerebras/MiniMax-M2-REAP-139B-A10B use?
Hugging Face lists the license as other. 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 this model on 8 GB VRAM?
Yes, using the IQ1_S quantization which requires approximately 31.6 GB of VRAM at 8k context.
Does the model support vision tasks?
No, the model does not include vision capabilities.
Is the model compatible with vLLM?
Yes, it works with vanilla vLLM without source modifications.