BenchLeader

Mistral Large 4.0 OSS vs Qwen3.5-35B-A3B

Full profiles: Mistral Large 4.0 OSS (Mistral AI) and Qwen3.5-35B-A3B (Alibaba): every benchmark, price and rank.

Updated · Data from 5 sources

As of 16 Nov 2026, there are not enough shared benchmarks for a verdict. Mistral Large 4.0 OSS leads in every category we measure both in.

Key facts (as of 16 Nov 2026)

  • Qwen3.5-35B-A3B costs $0.688 per 1M tokens blended (list price, 16 Nov 2026).
  • Of the 2 benchmarks both models have, Mistral Large 4.0 OSS leads 2 (BenchLeader, 16 Nov 2026).

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
MathsMistral Large 4.0 OSS66 vs 47 category score; OTIS Mock AIME 99.4% vs 54.4%High
ReasoningMistral Large 4.0 OSS64 vs 59 category score; GPQA Diamond 90.2% vs 83.5%Medium
Chat quality (human preference)–Only Qwen3.5-35B-A3B has an Arena rating (56 category score)Not comparable
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • Not enough shared benchmarks for a verdict.
  • Mistral Large 4.0 OSS (high) is stronger in maths, reasoning.

At a glance

SpecMistral Large 4.0 OSSQwen3.5-35B-A3B
DeveloperMistral AIAlibaba
Released29 Oct 202624 Feb 2026
Context window (tokens)–262,144
Compared atMistral Large 4.0 OSS (high)Qwen3.5-35B-A3B
Open weightsNoYes
Price in / out / cached input per 1M–$0.250 / $2.00 / $0.050

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryMistral Large 4.0 OSS (high)Qwen3.5-35B-A3BLeader
Reasoning64.258.7Mistral Large 4.0 OSS
Coding–37.3not comparable
Maths66.447.0Mistral Large 4.0 OSS
Knowledge34.1–not comparable
Human preference–56.4not comparable

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingMathsKnowledgeHumanpreferenceMistral Large 4.0 OSS, Reasoning: 64.2Mistral Large 4.0 OSS, Maths: 66.4Mistral Large 4.0 OSS, Knowledge: 34.1Qwen3.5-35B-A3B, Reasoning: 58.7Qwen3.5-35B-A3B, Coding: 37.3Qwen3.5-35B-A3B, Maths: 47.0Qwen3.5-35B-A3B, Human preference: 56.4

Dashed: that model has no result on the axis it crosses; the line only joins its neighbours.

Not drawn, no result for these models: Agents & tools, Instruction following, Long context, Multimodal.

Model, click to add or removeIndex

Where sources disagree

Across the 2 benchmarks both have, Mistral Large 4.0 OSS leads 2, Qwen3.5-35B-A3B leads 0.

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureMistral Large 4.0 OSS (high)Qwen3.5-35B-A3B
Blended price per 1M tokens–$0.688
Input / output per 1M– / –$0.250 / $2.00
Cached input per 1M–$0.050
Price per task: chat reply–$0.0007
Price per task: summarize a 30-page report–$0.0042
Price per task: code edit–$0.0045
Price per task: agentic coding session–$0.023
Price per task: structured extraction–$0.0009
Agentic coding session, input mostly cached–$0.014
Output speed–27 tok/s
Time to first answer–0.5 s

Each model is compared at its best-ranked reasoning-effort setting. Mistral Large 4.0 OSS (high). Qwen3.5-35B-A3B scores 49.9 and takes 1 s to its first answer token; Qwen3.5-35B-A3B (no reasoning) scores 47.5 in 4 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

41 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherMistral Large 4.0 OSS (high)Qwen3.5-35B-A3B
BenchLeader IndexBenchLeader–49.9
Knowledge scoreBenchLeader34.1–
Maths scoreBenchLeader66.447.0
Reasoning scoreBenchLeader64.258.7
Coding scoreBenchLeader–37.3
Human preference scoreBenchLeader–56.4
Blended price $/MBenchLeader–$0.688
Output speedBenchLeader–27 tok/s
Time to first answerBenchLeader–0.5 s
Context windowBenchLeader–262,144 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub90.2%83.5%
Arena Hard PromptsArena (LMArena)–1412
GPQA Diamond (AA)Artificial Analysis–81.9%†
Humanity's Last Exam (AA)Artificial Analysis–13.4%†
CritPtArtificial Analysis–0.6%†
Arena ExpertArena (LMArena)–1420
Chess PuzzlesEpoch AI Benchmarking Hub29.0%–
Mystery Game PuzzlesEpoch AI Benchmarking Hub20.0%–
LMCAEpoch AI Benchmarking Hub–29.5%
DTBenchEpoch AI Benchmarking Hub–80.0%
Coding
SciCodeSciCode–29.3%†
Arena CodingArena (LMArena)–1435
Arena WebDevArena (LMArena)–1253
ALE-BenchEpoch AI Benchmarking Hub–221.8
Terminal-Bench 2.1 (AA)Artificial Analysis–40.8%†
Agents & tools
Terminal-Bench HardArtificial Analysis–10.6%†
τ²-Bench Telecom (AA)Artificial Analysis–86.3%†
GDPval-AA v2.1Artificial Analysis–616†
τ³-Banking (AA)Artificial Analysis–5.0%†
ITBench SRE (AA)Artificial Analysis–21.5%†
Maths
OTIS Mock AIMEEpoch AI Benchmarking Hub99.4%54.4%
AIME 2026MathArena–93.3%
HMMT February 2026MathArena–81.8%
MathArena ApexMathArena–4.2%
Arena MathsArena (LMArena)–1397
Show all 41 benchmarks
MetricPublisherMistral Large 4.0 OSS (high)Qwen3.5-35B-A3B
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub22.4%–
EBR-benchEpoch AI Benchmarking Hub18.1%–
Global-MMLU-Lite (AA)Artificial Analysis–82.4%†
Instruction following
IFBenchArtificial Analysis–44.5%†
Arena Instruction FollowingArena (LMArena)–1387
Human preference
Arena TextArena (LMArena)–1394
Arena Creative WritingArena (LMArena)–1342
Arena Multi-turnArena (LMArena)–1394
Arena Longer QueriesArena (LMArena)–1399
Multimodal
MMMU-ProArtificial Analysis–69.2%†
Furniture AssemblyEpoch AI Benchmarking Hub30.8%–
Long context
AA-LCR v1.1Artificial Analysis–63.0%†
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub–144.0
AA Intelligence Index v4.3.2Artificial Analysis–15.1†
AA Progress Index v1Artificial Analysis–238.1†
Metadata
AA Openness IndexArtificial Analysis–38.9†

Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.

What changed

  • Qwen3.5-35B-A3B: Qwen3.5-35B-A3B cached input price moved from $0.15/M to $0.050/M (67% down).
  • Qwen3.5-35B-A3B: BenchLeader Index 50.4 → 49.9.
  • Qwen3.5-35B-A3B: 1 new result (ALE-Bench).
  • Qwen3.5-35B-A3B: 1 score revision.
  • Mistral Large 4.0 OSS (high): first results.
  • Qwen3.5-35B-A3B: Qwen3.5-35B-A3B cached input price moved from $0.05/M to $0.15/M (200% up).

Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs

Mistral Large 4.0 OSS vs Qwen3.5-35B-A3B: questions

Do independent benchmarks agree on Mistral Large 4.0 OSS vs Qwen3.5-35B-A3B?
Across the 2 benchmarks both have, Mistral Large 4.0 OSS leads 2, Qwen3.5-35B-A3B leads 0.
When was Mistral Large 4.0 OSS released?
Mistral Large 4.0 OSS was released on 29 Oct 2026.