Mistral AI · Devstral

Devstral Small 2505

READER SUMMARY

Mistral and All Hands AI's Apache-licensed 24B agentic software-engineering model for repository-scale tasks.

DOOM SCORE75.3out of 100model risk profile, not the overall index

CURRENT ASSESSMENT · REVISION 1

Why this model scores 75.3

Agent-scaffold optimization and downloadable weights enable autonomous code changes in private deployments. Smaller size limits frontier capability while open licensing raises misuse and control difficulty.

Capability69

Autonomy80

Deployment92

Misuse potential66

Control difficulty72

0 comments · 0 votesOpen discussion

No comments yet. Start the discussion.

MODEL-ATTRIBUTED EVIDENCE

News tied to Devstral Small 2505

The model score of 75.3 rates this model's risk profile. The overall Doom Index of 63.5 measures the recalibrated complete evidence corpus. These values answer different questions.

NET MODEL-ATTRIBUTED INDEX CONTRIBUTION+0.03

Each article's leave-one-out Doom Index contribution is divided equally among the exact models named on that article. This prevents multi-model evidence from being claimed in full on several model pages. Model risk scores are recalculated from their technical profile and related news, but never feed back into the overall index.

21 May 2025AutonomyTOWARD

Mistral and All Hands release open agentic coding model Devstral

Mistral and All Hands AI released Devstral Small 2505 under Apache 2.0 for repository-scale software engineering through agent scaffolds such as OpenHands and SWE-Agent.

Full item contribution: +0.03
Devstral Small 2505 equal share: +0.03

Read assessment →

AUDIT TRAIL

Model score history

R1
Doom Score 75.3
Initial exact profile for Devstral Small 2505.
12 Aug 2026