Devstral Medium

Name: Devstral Medium benchmark results
Creator: BenchLeader

Best configuration ranks #511 of 610 on the BenchLeader Index at 40.3 ±8.3 (medium reasoning effort). Last measured 8 Sept 2026. Released 10 Jul 2025.

Blended price: $0.800/M; $0.400 in · $2.00 out
Output speed: –
First answer: –
Context: 128k

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

Overall index40
Reasoning36
Coding11
Agents & tools42
Knowledge49
Instruction following34
Long context41
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	medium	Source
GPQA Diamond (AA)not in index	49.2%#409	Artificial Analysis
Humanity's Last Exam (AA)not in index	3.8%#486	Artificial Analysis
Kagi LLM Benchmark	37.7%#112	Kagi LLM Benchmark

Coding

Benchmark	medium	Source	Trend
LMArena WebDev	1081#118	LMArena

Agents & tools

Benchmark	medium	Source	Trend
Terminal-Bench Hard	9.1%#223	Artificial Analysis
τ²-Bench Telecom (AA)not in index	19.9%#321	Artificial Analysis

Knowledge

Benchmark	medium	Source	Trend
AA-Omniscience	-31.6#221	Artificial Analysis

Instruction following

Benchmark	medium	Source	Trend
IFBench	29.9%#346	Artificial Analysis

Long context

Benchmark	medium	Source	Trend
AA-LCR	32.0%#315	Artificial Analysis

Composite

Benchmark	medium	Source	Trend
AA Intelligence Index	9.0#340	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	$0.0008	–	–
Summarise a 30-page report	12,000 / 600	$0.0060	–	–
Code edit	6,000 / 1,500	$0.0054	–	–
Agentic coding session	60,000 / 4,000	$0.032	–	–
Structured extraction	2,000 / 200	$0.0012	–	–