Devstral Small 2 24BvsQwen 2.5 Coder 32B
Complete comparison between Devstral Small 2 24B (24B parameters, Mistral AI) and Qwen 2.5 Coder 32B (32B, Alibaba). VRAM required by quantization, tokens per second measured on 4 reference GPUs, verdict by use case, license, installation commands. All figures are calculated from catalog data — no copy-pasting between pages.
In brief
| Feature | Devstral Small 2 24B | Qwen 2.5 Coder 32B |
|---|---|---|
| Settings | 24B | 32B |
| Family | Mistral | Qwen |
| Author | Mistral AI | Alibaba |
| Origin | FR | CN |
| License | Apache 2.0 | Apache 2.0 |
| Context | 256,000 tokens | 131,072 tokens |
| Output | December 2025 | November 2024 |
Memory footprint
Approximate VRAM required for inference with an average context window. The winner (in green) is the model that uses less — smaller models have the edge.
| Quantization | Devstral Small 2 24B | Qwen 2.5 Coder 32B |
|---|---|---|
| Q4_K_M (lightweight) | 14 GB | 19 GB |
| Q5_K_M (balanced) | 17 GB | 23 GB |
| Q8 (near-lossless) | 26 GB | 35 GB |
| FP16 (max quality) | 48 GB | 64 GB |
| CPU-only RAM | 24 GB | 32 GB |
Estimated speed (tokens/second)
Estimates based on the best quantization that can fit on each GPU. Actual figures depend on the prompt, context, and engine (llama.cpp, vLLM, MLX). Methodology.
| Reference GPU | Devstral Small 2 24B | Qwen 2.5 Coder 32B |
|---|---|---|
| RTX 4090 (24 GB) | 40 tok/s · Q5_K_M | 30 tok/s · Q5_K_M |
| RTX 4080 (16 GB) | 15 tok/s · Q4_K_M | ✗ too heavy |
| RTX 3060 12GB (12 GB) | ✗ too heavy | ✗ too heavy |
| Apple M4 Pro (48 GB) (36 GB) | 15 tok/s · Q8 | 12 tok/s · Q8 |
Verdict by use case
For each common use case, we indicate which of the two is better suited based on its tags, size, and specialization.
Replace GitHub Copilot and Cursor with a 100% local coding copilot—the definitive guide, configs included.
- Lifetime online access
- PDF + files
- 30-day money-back guarantee
Strengths and weaknesses
Devstral Small 2 24B
24B coding specialist, Apache 2.0. 72.2% SWE-Bench. 256k ctx, FR lab.
- 68.0% SWE-bench Verified
- Fits on RTX 4090
- Apache 2.0
- French lab
- No vision
- Dense: slower than code MoEs (Qwen3-Coder 30B-A3B) at the same memory capacity
Installation
Qwen 2.5 Coder 32B
The open-weight code reference from late 2024, now surpassed by the Qwen3-Coder / Devstral generation.
- Best open-weight coding LLM
- Very strong on HumanEval (2024 benchmark)
- 128k context
- Requires 20+ GB of VRAM
- Worse at general chat
Installation
Frequently asked questions
Which model runs best on RTX 4090 (24 GB): Devstral Small 2 24B or Qwen 2.5 Coder 32B?
On a RTX 4090, Devstral Small 2 24B runs at Q5_K_M (~40 tok/s), Qwen 2.5 Coder 32B in Q5_K_M (~30 tok/s). In pure throughput, Devstral Small 2 24B wins. See the configurator to test your exact GPU.
Which uses the least VRAM?
In Q4_K_M, Devstral Small 2 24B fits in 14 GB versus 19 GB for its rival—a 5 GB difference, significant if you're targeting an RTX 3060 or a 4060 Ti 8 GB.
Devstral Small 2 24B or Qwen 2.5 Coder 32B for coding?
Both are specialized code models—the choice depends on context (256,000 vs. 131,072 tokens), licensing (Apache 2.0 vs. Apache 2.0), and available VRAM. Both integrate with Continue or Aider.
Can these models be used in commercial production?
Devstral Small 2 24B is under Apache 2.0 — free for commercial use. Qwen 2.5 Coder 32B is under Apache 2.0 — also free to use. For SaaS, prefer Apache 2.0 or MIT.
Which one should you choose in 2026?
Depends on your primary constraint. Smaller / faster: Devstral Small 2 24B (24B). More capable: Qwen 2.5 Coder 32B (32B). If you're unsure, run it configurator with your GPU and use case — it will decide based on both.
Alternatives to consider
If neither works for you, here are some neighboring models you could explore.