Devstral Small (May '25) vs Mistral Medium 3.5

Wins 0 of 5 benchmarks
$0.50 / task
Wins 1 of 5 benchmarks
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Devstral Small (May '25)4
Mistral Medium 3.513.8
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
Devstral Small (May '25)—
Mistral Medium 3.50
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Devstral Small (May '25)—
Mistral Medium 3.56.3
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
Devstral Small (May '25)—
Mistral Medium 3.51,400
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
Devstral Small (May '25)—
Mistral Medium 3.564.9

Specs

SpecDevstral Small (May '25)Mistral Medium 3.5
Input, per 1M!$0.06$1.50
Output, per 1M!$0.12$7.50
Cost per task?—$0.50
Speed, tok/s—155
Context256K262K
WeightsOpenOpen
ReleasedMay 2025Apr 2026
Memory (4-bit)—77 GB
Licence—custom licence
Commercial use—Restricted
See images?✓
Read PDFs?✓
Hear audio?—
Watch video?—
Call tools✓✓
Force a tool call?Partial
Return strict JSON—✓
Search the web??
Run code??
Use a computer??

What breaks when you switch

Devstral Small (May '25) → Mistral Medium 3.5
Nothing known.
Mistral Medium 3.5 → Devstral Small (May '25)
Context: 262K → 256K
Memory (4-bit): not verified
No strict JSON