Mistral AIOpen weightsAuto-detected

Devstral Small 2

Name: Devstral Small 2 benchmark results
Creator: BenchLeader

Best configuration ranks #474 of 610 on the BenchLeader Index at 41.9 ±4.7. Released 9 Dec 2025.

Blended price: –
Output speed: 136 tok/s
First answer: 2.28 s; first token 2.28 s
Context: 256k

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index42
Agents & tools48
Knowledge37
Instruction following35
Multimodal28
Long context39
Composite40

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	default	Source
GPQA Diamond (AA)not in index	53.2%#392	Artificial Analysis
Humanity's Last Exam (AA)not in index	3.5%#508	Artificial Analysis

Coding

Benchmark	default	Source
SciCode (AA)not in index	32.4%#141	Artificial Analysis

Agents & tools

Benchmark	default	Source
Terminal-Bench Hard	16.7%#179	Artificial Analysis
τ²-Bench Telecom (AA)not in index	23.4%#291	Artificial Analysis

Knowledge

Benchmark	default	Source
AA-Omniscience	-57.7#371	Artificial Analysis

Instruction following

Benchmark	default	Source
IFBench	31.2%#339	Artificial Analysis

Multimodal

Benchmark	default	Source
MMMU-Pro	44.6%#219	Artificial Analysis

Long context

Benchmark	default	Source
AA-LCR	28.0%#328	Artificial Analysis

Composite

Benchmark	default	Source
AA Intelligence Index	8.4#364	Artificial Analysis

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	–	–	4.5 s
Summarise a 30-page report	12,000 / 600	–	–	6.7 s
Code edit	6,000 / 1,500	–	–	13.3 s
Agentic coding session	60,000 / 4,000	–	–	31.8 s
Structured extraction	2,000 / 200	–	–	3.8 s