Inference
EN
Alibaba

Qwen3 Max

Qwen3 Max is a reasoning model from Alibaba. 20 benchmarks count toward its index, in 6 categories.

  • Closed weights
  • Reasoning
  • Input: text
  • Released 2025-09
  • Retires 9 Oct 2026
RankedEnough results for an index.
IndexAll categories, out of 100.45.0±5.1Median 51.0
CoverageIndex weight that has results.85%
SpeedOutput tokens per second.62/sMedian 67/s
First tokenSeconds to the first output token.0.7 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.78Median $0.60
Output / 1MUS dollars per 1M output tokens.$3.90Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.1413±6 #138Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
qwen/qwen3-max
Max outputMaximum output tokens per request.
66K
Knowledge cutoffLast date of training data.
30 Jun 2025
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.156
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
9 Oct 2026

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.4 counted47.1153 / 218
CodingCode writing and repair.3 counted46.3158 / 209
ReasoningLogic problems and puzzles.3 counted40.3161 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.5 counted48.1149 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted30.2—
MathMath problems.4 counted46.3101 / 168

Results

20 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
MATH level 5Math97.1% ±0.46 / 88
SimpleQA VerifiedKnowledgeAbove fitFar above this model's index.48.7% ±1.633 / 91
IOI v1Coding15.7% ±6.428 / 61
MMLU ProKnowledge84.4% ±0.462 / 135
OTIS Mock AIME 2024-2025Math73.3% ±5.994 / 200
AIMEMath81.0% ±1.146 / 95
GPQA diamond · Epoch AIKnowledge72.6% ±2.7110 / 217
τ²-Bench Tool-Agent-User EvaluationAgentic74.3%76 / 149
GPQA Diamond · Vals AIKnowledge79.5% ±2.069 / 135
MMLU ProKnowledge83.5% ±0.471 / 135
LiveCodeBenchCoding78.2% ±1.176 / 142
GPQA Diamond · Vals AIKnowledge77.8% ±2.175 / 135
Artificial Analysis GPQA DiamondKnowledge76.4%112 / 199
IOI v1Coding7.8% ±3.236 / 61
Terminal-Bench 1.0Agentic36.3% ±5.428 / 47
Terminal-Bench 1.0Agentic36.3% ±5.428 / 47
Gert Labs Composite Game BenchmarkAgentic43.7%31 / 52
AIMEMath60.7% ±2.657 / 95
Artificial Analysis IFBenchInstruction44.1%89 / 146
Artificial Analysis Humanity's Last ExamKnowledge11.9%136 / 222
Artificial Analysis Omniscience AccuracyKnowledge24.4%107 / 164
τ²-bench AirlineAgentic59.5%15 / 22
LiveCodeBenchCoding66.9% ±1.297 / 142
Critical Physics TasksReasoning0.0%150 / 214
Artificial Analysis Long Context ReasoningReasoning50.0%154 / 217
τ²-bench TelecomAgentic84.2%13 / 18
Terminal-Bench 2.0 · Vals AIAgentic24.7% ±4.652 / 66
τ²-bench RetailAgentic72.2%19 / 23
FrontierMath-Tiers-1-3-v2-PrivateMath18.9% ±2.380 / 90
Vibe Code Bench v1.1Coding3.5% ±1.3113 / 123
Mystery Game PuzzlesReasoning5.0% ±2.275 / 80
MGSMMultilingual92.1% ±1.621 / 74
TaxEval v2Finance74.0% ±0.942 / 143
TaxEval v2Finance73.5% ±0.948 / 143
MGSMMultilingual91.8% ±1.725 / 74
LegalBenchLegal81.9% ±0.478 / 149
LegalBenchLegal80.3% ±0.487 / 149
MedQAHealthcare87.4% ±0.357 / 95
MedQAHealthcare87.4% ±0.357 / 95
CorpFin v2Finance55.9% ±1.091 / 131
CorpFin v2Finance53.0% ±1.0100 / 131
Finance Agent v1.1Finance44.3% ±2.839 / 51
MedScribeHealthcare72.7% ±1.987 / 109
MedCodeHealthcare31.4% ±1.993 / 106
CaseLaw v2Legal47.5% ±0.551 / 53

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYVals AINo licence statedSierra ResearchMIT

History

5 releases
Index46.0 → 45.0
23 Oct 202612 Nov 2026
Output / 1M$3.90 → $3.90
23 Oct 202612 Nov 2026

Questions

How much does Qwen3 Max cost?
$0.78 per 1M input tokens, $3.90 per 1M output tokens.
How good is Qwen3 Max?
Its index is 45.0 out of 100. That is place 178 of 283 ranked models.
What is Qwen3 Max best at?
Knowledge: 48.1 out of 100, place 149 of 268.
How fast is Qwen3 Max?
62 output tokens per second, 0.7 s to the first token.
What is the context window of Qwen3 Max?
262K tokens.
Is Qwen3 Max open source?
No. Its weights are not published.
When was Qwen3 Max released?
September 2025.

Badge

Qwen3 Max on Inference360: place 178 of 283, index 45.0
HTML
<a href="https://inference360.ai/models/qwen3-max"><img src="https://inference360.ai/badges/qwen3-max.svg" alt="Qwen3 Max on Inference360: place 178 of 283, index 45.0" height="20"></a>
Markdown
[![Qwen3 Max on Inference360: place 178 of 283, index 45.0](https://inference360.ai/badges/qwen3-max.svg)](https://inference360.ai/models/qwen3-max)

Compare with

Grok 4 Fast (Reasoning)45.3o4-mini44.7GLM-4.545.7Qwen3.5 397B (Reasoning)55.3

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeNemotron 3 Ultra50.8 · Free

More from Alibaba

All models
Qwen3 VL 235B A22B ThinkingUnrankedQwen3 VL 235B A22B InstructUnrankedQwen3 Coder PlusUnrankedQwen3-Omni-30B-A3B-InstructUnrankedQwen3-Omni-30B-A3B-ThinkingUnranked