Model
Model explorer

Devstral Small

OPEN
Mistral AI · Devstral family · released May 21, 2025

24B open agentic-coding model built with All Hands AI; 46.8% SWE-bench Verified, best open model under 100B at launch.

ReasoningCodingVisionFunction callingTool useAgentic
1330.2
Elo · unrated
Parameters
24B
Active params
24B (dense)
Context
128K tokens
Architecture
Dense transformer — 24B, finetuned from Mistral-Small-3.1-24B-Base
License
Apache 2.0
Languages
—
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
SWE-bench VerifiedCoding46.8%#100
best: Claude Opus 5 · 96.0%
Run it locally
VRAM @ Q4
14.3 GB
VRAM @ FP16
47.2 GB
Fits on (Q4)
RTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
GGUF Q4 · AWQ · MLX
Fine-tune it
Permissive
QLoRA17.1 GB1× RTX 3090 24GB
LoRA51.9 GB1× A100 80GB
Full fine-tune386.0 GB4× H200 141GB
QLoRA SFT on ~10k samples ≈ $12.36 (1× RTX 3090 24GB)
Plan a fine-tune →
Devstral family
Elo progression across releases
Devstral SmallMay 20251330.2
Resources
HFmistralai/Devstral-Small-2505GHmistralai/mistral-inferenceDOCSdocs.mistral.ai/models/model-cards/devstral-small-1-0-25-05
API price weights · each benchmark row carries its own source badge (see methodology)