What are Mistral Large 4's benchmark scores?

As of 25 Oct 2026, Mistral Large 4 scores 56.7 ±5.5 on the BenchLeader Index, strongest in long context (66) and weakest in composite (50). That rests on 18 independent benchmark results, and every score on this page links to its publisher.

How does Mistral Large 4 rank?

Mistral Large 4 is #196 of 785 ranked configurations on 25 Oct 2026. 288 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.

How much does Mistral Large 4 cost per 1M tokens?

As of 25 Oct 2026, Mistral Large 4 costs $1.36 per 1M input tokens and $4.18 per 1M output tokens, $2.06 blended (three input tokens per output token).

When was Mistral Large 4 released?

Mistral AI released Mistral Large 4 on 6 Oct 2026. BenchLeader has results for it at 2 reasoning-effort settings.

Mistral AIOpen weightsReasoning modelCompare with another model

Mistral Large 4

Reasoning effort

Mistral Large 4 is a Mistral AI open-weights reasoning model, released 6 Oct 2026. Its best configuration ranks #196 of 785 on the BenchLeader Index at 56.7 ±5.5, in the upper half. The ± is the point: 288 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in long context (66) and lowest in composite (50). At $2.06 per million tokens blended it is pricier than most ranked models. Output speed of 95 tokens per second puts it faster than most, with a first answer in 24.3 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 23 Oct 2026.

Blended price: $2.06/M; $1.36 in · $4.18 out
Output speed: 95 tok/s; measured by Artificial Analysis
First answer: 24 s; first token 1.70 s
Context: 1.0M
Full answer: 30 s; median, reasoning included
Cost per run: $1.13; one full Intelligence Index run
Released: 6 Oct 2026

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

Overall index57
Reasoning59
Coding60
Agents & tools50
Maths55
Knowledge59
Human preference62
Multimodal60
Long context66
Composite50

Versions

Mistral AI has shipped 5 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
Mistral Large 4this page	6 Oct 2026	56.7	#196
Mistral Large 3	2 Dec 2025	45.4	#512
Mistral Large 2.1	18 Nov 2024	42.5	#590
Mistral Large 2	24 Jul 2024	40.7	#652
Mistral Large 24.02	26 Feb 2024	40.2	#660

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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
high	49.8	#386	95 tok/s	24 s	$0.0018	Coding 48 · Knowledge 32 · Maths 56 · Reasoning 63
not statedbest	56.7	#196	95 tok/s	24 s	$0.0018	Agents & tools 50 · Coding 60 · Composite 50 · Human preference 62 · Knowledge 59 · Long context 66 · Maths 55 · Multimodal 60 · Reasoning 59

How its index has moved

6 Oct 2026 to 25 Oct 2026

not stated
high

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. 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	high	not stated	Source
GPQA Diamond	88.1%#64	–	Epoch AI Benchmarking Hub
Arena Hard Promptseffort not stated	–	1456#105	Arena (LMArena)
LiveBench 2.0 Reasoningnot in indexeffort not stated	–	83.9%#33	LiveBench
LiveBench 1.0 Reasoningnot in index	83.9%#42	–	LiveBench
Humanity's Last Exam (AA)not in indexeffort not stated	–	35.0%#121	Artificial Analysis
CritPteffort not stated	–	10.6%#117	Artificial Analysis
MysteryMechanismnot in index	12.2%#28	–	Vals AI
Chess Puzzlesnot in index	33.0%#61	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	17.0%#75	–	Epoch AI Benchmarking Hub
LMCAnot in indexeffort not stated	–	35.8%#124	Epoch AI Benchmarking Hub
DTBenchnot in indexeffort not stated	–	88.2%#95	Epoch AI Benchmarking Hub

Coding

Benchmark	high	not stated	Source
SciCodeeffort not stated	–	54.2%#64	SciCode
FrontierCode	39.4%#28	–	Cognition
Arena Codingeffort not stated	–	1498#83	Arena (LMArena)
Arena WebDeveffort not stated	–	1545#44	Arena (LMArena)
LiveBench 2.0 Codingnot in indexeffort not stated	–	60.0%#46	LiveBench
LiveBench 1.0 Codingnot in index	77.2%#39	–	LiveBench
SciCode (AA)not in indexeffort not stated	–	54.2%#67	Artificial Analysis
IOI	45.3%#30	–	Vals AI
Code Migrationnot in index	30.6%#41	–	Vals AI
SWE-bench Multimodalnot in index	33.1%#20	–	Vals AI
Vibe Code Bench 1-100not in index	14.3%#21	–	Vals AI
Vibe Code Bench v1.1not in index	78.4%#30	–	Vals AI
Surface Evolver Benchnot in indexeffort not stated	–	28.1%#25	Epoch AI Benchmarking Hub
ALE-Benchnot in indexeffort not stated	–	1213.2#40	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in indexeffort not stated	–	26.8%#49	Artificial Analysis
FrontierCode 1.1 (AA)not in indexeffort not stated	–	41.5%#29	Artificial Analysis

Agents & tools

Benchmark	high	not stated	Source
Arena Agenteffort not stated	–	-5.1#36	Arena (LMArena)
LiveBench Long Agenticnot in indexeffort not stated	–	11.8%#48	LiveBench
LiveBench Browsernot in indexeffort not stated	–	8.0%#42	LiveBench
LiveBench Agentic Codingnot in index	57.2%#22	–	LiveBench
GDPval-AA v2.1not in indexeffort not stated	–	46.1%#48	Artificial Analysis
Finance Agent v2not in index	54.7%#26	–	Vals AI
Finance Agent v2.1not in index	61.4%#15	–	Vals AI
Harvey's Legal Agent Benchmark v1.1not in index	17.5%#6	–	Vals AI
Harvey's Legal Agent Benchmark v1not in index	15.8%#10	–	Vals AI
Legal Research Benchnot in index	31.7%#46	–	Vals AI
Tax Agent Benchnot in index	30.4%#27	–	Vals AI
Terminal-Bench 4.0 (Vals)not in index	27.8%#23	–	Vals AI
AutomationBenchnot in indexeffort not stated	–	59.9%#38	Artificial Analysis
GDP.pdf (AA)not in indexeffort not stated	–	18.6%#71	Artificial Analysis
Harvey LAB-AA v1.1not in indexeffort not stated	–	3.1%#13	Harvey
AA-Briefcase v1.1not in indexeffort not stated	–	1391#30	Artificial Analysis
Terminal-Bench-Science 0.1 (AA)not in indexeffort not stated	–	4.8%#49	Artificial Analysis
CWE-Bench (AA)effort not stated	–	50.8%#15	Artificial Analysis
DeepSec-Bench (AA)effort not stated	–	16.0%#14	Artificial Analysis
CyberGym end-to-end (AA)effort not stated	–	81.7%#2	Artificial Analysis

Maths

Benchmark	high	not stated	Source
FrontierMath Tiers 1–3	58.6%#43	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	26.8%#43	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	98.3%#24	–	Epoch AI Benchmarking Hub
ProofBencheffort not stated	–	10.0%#65	Vals AI
LiveBench Mathematicsnot in index	93.6%#21	–	LiveBench
Arena Mathseffort not stated	–	1467#55	Arena (LMArena)

Knowledge

Benchmark	high	not stated	Source
SimpleQA Verified	20.0%#74	–	Epoch AI Benchmarking Hub
LiveBench 2.0 Data Analysisnot in indexeffort not stated	–	76.5%#23	LiveBench
LiveBench Researchnot in indexeffort not stated	–	28.1%#52	LiveBench
LiveBench 1.0 Data Analysisnot in index	76.5%#32	–	LiveBench
AA-Omniscienceeffort not stated	–	-5.3#165	Artificial Analysis
BioMysteryBenchnot in index	67.0%#17	–	Vals AI
Excel Modeling Benchmarknot in index	56.2%#47	–	Vals AI
Excel Modeling Benchmark v1.1not in index	69.0%#18	–	Vals AI
MedCodenot in index	40.7%#59	–	Vals AI
MedScribenot in index	80.4%#55	–	Vals AI
AA-Omniscience: accuracynot in indexeffort not stated	–	25.8%#223	Artificial Analysis
AA-Omniscience: non-hallucinationnot in indexeffort not stated	–	58.0%#84	Artificial Analysis

Instruction following

Benchmark	high	not stated	Source
LiveBench Languagenot in index	49.6%#66	–	LiveBench
LiveBench Instruction Followingnot in index	64.8%#45	–	LiveBench
Arena Instruction Followingnot in indexeffort not stated	–	1426#101	Arena (LMArena)

Human preference

Benchmark	not stated	Source
Arena Text	1434#106	Arena (LMArena)
Arena Creative Writingnot in index	1369#149	Arena (LMArena)
Arena Multi-turnnot in index	1443#98	Arena (LMArena)
Arena Longer Queriesnot in index	1436#111	Arena (LMArena)

Multimodal

Benchmark	not stated	Source
Arena Vision	1272#42	Arena (LMArena)
LiveBench Designnot in index	65.6%#43	LiveBench
MMMU-Pro	76.4%#88	Artificial Analysis
AA-3Dnot in index	870#18	Artificial Analysis

Long context

Benchmark	not stated	Source	Trend
AA-LCR	81.3%#56	Artificial Analysis
MLCRnot in index	7.8%#129	Artificial Analysis

Composite

Benchmark	high	not stated	Source
Epoch Capabilities Indexnot in indexeffort not stated	–	150.0#59	Epoch AI Benchmarking Hub
LiveBench 2.0effort not stated	–	47.7%#48	LiveBench
LiveBench 1.0not in index	71.8%#52	–	LiveBench
AA Intelligence Index v4.3.2effort not stated	–	38.4#76	Artificial Analysis
Vals Finance Indexnot in index	56.9%#16	–	Vals AI
AA Cyber Indexnot in indexeffort not stated	–	49.5#7	Artificial Analysis

Safety & honesty

Benchmark	not stated	Source	Trend
FORTRESSnot in index	23.1%#34	Scale AI SEAL

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

#2CyberGym end-to-end (AA)81.7% · of 21

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.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Mistral	106 tok/s	1.20 s	$1.36	$4.18	1.0M	–
Mistral	100 tok/s	0.78 s	$1.36	$4.18	1.0M	–

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 at $0.140 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0018	$0.0014	27.5 s
Summarize a 30-page report	12,000 / 600	$0.019	$0.0078	30.6 s
Code edit	6,000 / 1,500	$0.014	$0.0089	40.2 s
Agentic coding session	60,000 / 4,000	$0.098	$0.043	1.1 min
Structured extraction	2,000 / 200	$0.0036	$0.0017	26.4 s

New this month

Questions people ask

What are Mistral Large 4's benchmark scores?: As of 25 Oct 2026, Mistral Large 4 scores 56.7 ±5.5 on the BenchLeader Index, strongest in long context (66) and weakest in composite (50). That rests on 18 independent benchmark results, and every score on this page links to its publisher.
How does Mistral Large 4 rank?: Mistral Large 4 is #196 of 785 ranked configurations on 25 Oct 2026. 288 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
How much does Mistral Large 4 cost per 1M tokens?: As of 25 Oct 2026, Mistral Large 4 costs $1.36 per 1M input tokens and $4.18 per 1M output tokens, $2.06 blended (three input tokens per output token).
When was Mistral Large 4 released?: Mistral AI released Mistral Large 4 on 6 Oct 2026. BenchLeader has results for it at 2 reasoning-effort settings.

Data as of 25 Oct 2026. Compare these configurations.

Cite as: BenchLeader, “Mistral Large 4: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/mistral-large-4.