Devstral 2 (123B)
Mistral AI's flagship coding agent at 123B parameters — SOTA on SWE-bench Verified. Same Ministral 3 architecture with vision support and 256K context. Requires 128GB RAM/VRAM.
Model Specifications
Estimated Quantization Sizes
| Format | Precision | Est. VRAM | Recommendation |
|---|---|---|---|
| FP16 / BF16 | 16-bit | 246.0 GB | Uncompressed Base |
| Q8_0High | 8-bit | 123.0 GB | Near Lossless |
| Q6_K | 6-bit | 92.3 GB | Excellent Balance |
| Q4_K_MPopular | 4-bit | 61.5 GB | Standard Use |
Share this Model
Send this model's specs directly to your community.
Similar Models
Devstral Small 22B
22BMistral's open-source agentic coding model optimized for SWE tasks and codebase exploration, state-of-the-art on SWE-Bench
Devstral Small 2
24BMistral AI's updated coding agent — 24B parameters, Apache 2.0, 256K context window with vision support. Mistral Vibe CLI included. Built on Ministral 3 architecture with Scalable-Softmax.
Devstral Medium
0BCoding-focused enterprise model from Mistral. Achieves 61.6% on SWE-bench Verified, placing it ahead of GPT-4.1 in code tasks. API-only; supports private infrastructure deployment and agentic coding frameworks.
Related Guides
How much VRAM do you really need?
A complete breakdown of quantization levels and VRAM overhead for running local models.
Best GPUs for Machine Learning in 2026
Comparing NVIDIA and AMD options for the best speed-to-dollar ratio.
GGUF vs EXL2 vs AWQ
Understanding local AI formats and which one to pick for your specific hardware.