BenchLeader
Menu
LeaderboardsModelsProvidersBenchmarksCompareReleasesBlogMethodology
Mistral AIAuto-detected

Mistral Medium 3.5

Best configuration ranks #277 of 610 on the BenchLeader Index at 50.8 ±7.7. Last measured 8 Sept 2026.

Blended price
$3.00/M
$1.50 in · $7.50 out
Output speed
137 tok/s
First answer
17 s
first token 0.26 s
Context
262k
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Overall index51
  2. Reasoning50
  3. Coding41
  4. Agents & tools45
  5. Knowledge47
  6. Instruction following67
  7. Human preference47
  8. Multimodal53
  9. Long context61
  10. Composite48

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.

EffortIndexRankSpeedFirst answerChat reply costCategories
high––137 tok/s17 s$0.0029Agents & tools 29 · Knowledge 45
defaultbest50.8#277137 tok/s17 s$0.0029Agents & tools 45 · Coding 41 · Composite 48 · Human preference 47 · Instruction following 67 · Knowledge 47 · Long context 61 · Multimodal 53 · Reasoning 50

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

BenchmarkhighdefaultSourceTrend
LMArena Hard Prompts–1446#99LMArena
GPQA Diamond (AA)not in index–74.8%#245Artificial Analysis
Humanity's Last Exam (AA)not in index–13.8%#206Artificial Analysis
GPQA Diamond (Vals)not in index34.9%#125–Vals AI
Kagi LLM Benchmark–41.4%#101Kagi LLM Benchmark

Coding

BenchmarkhighdefaultSourceTrend
LMArena Coding–1479#93LMArena
LMArena WebDev–1265#103LMArena
SciCode (AA)not in index–40.2%#114Artificial Analysis
SWE-bench (Vals)not in index66.4%#69–Vals AI

Agents & tools

BenchmarkhighdefaultSourceTrend
LMArena Agent–-9.3#37LMArena
Terminal-Bench Hard–33.3%#90Artificial Analysis
τ²-Bench Telecom (AA)not in index–94.2%#34Artificial Analysis
Terminal-Bench 2.1 (Vals)39.0%#57–Vals AI

Knowledge

BenchmarkhighdefaultSourceTrend
AA-Omniscience–-36.8#242Artificial Analysis
MMLU-Pro75.3%#105–Vals AI
CorpFin58.8%#80–Vals AI
TaxEval68.0%#100–Vals AI

Instruction following

BenchmarkhighdefaultSourceTrend
IFBench–68.8%#77Artificial Analysis

Human preference

BenchmarkhighdefaultSourceTrend
LMArena Text–1427#96LMArena
EQ-Bench 4–993#27EQ-Bench

Multimodal

BenchmarkhighdefaultSourceTrend
LMArena Vision–1222#65LMArena
MMMU-Pro–64.9%#147Artificial Analysis

Long context

BenchmarkhighdefaultSourceTrend
AA-LCR–69.3%#170Artificial Analysis

Composite

BenchmarkhighdefaultSourceTrend
Epoch Capabilities Indexnot in index–141.3#93Epoch AI Benchmarking Hub
AA Intelligence Index–14.9#224Artificial Analysis
Vals Indexnot in index17.9#49–Vals AI

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

ProviderSpeedFirst tokenInput $/MOutput $/MContextQuantisation
Mistral (ZDR)18 tok/s0.27 s$1.50$7.50262k–
Mistral17 tok/s0.26 s$1.50$7.50262k–

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.

WorkloadTokens in / outCostWith cachingTime
Chat reply400 / 300$0.0029–19.1 s
Summarise a 30-page report12,000 / 600$0.022–21.3 s
Code edit6,000 / 1,500$0.020–27.9 s
Agentic coding session60,000 / 4,000$0.120–46.2 s
Structured extraction2,000 / 200$0.0045–18.4 s

See also

Best for codingBest for agentsAll Mistral AI models

Data as of 9 Sept 2026. Compare these configurations.