Inference
EN
Alibaba

Qwen3.7 Max

Qwen3.7 Max is a reasoning model from Alibaba. 54 benchmarks count toward its index, in 7 categories.

  • Closed weights
  • Reasoning
  • Input: text
  • Released 2026-05
RankedEnough results for an index.
IndexAll categories, out of 100.60.5±3.2Median 51.0
CoverageIndex weight that has results.90%
SpeedOutput tokens per second.68/sMedian 67/s
First tokenSeconds to the first output token.1.3 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$1.48Median $0.60
Output / 1MUS dollars per 1M output tokens.$4.43Median $2.55
ContextMaximum tokens in one request.1MMedian 262K
EloLMArena rating and rank.1475±10 #26Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
qwen/qwen3.7-max
Max outputMaximum output tokens per request.
131K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.295
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
5 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Novita AIHost$1.25$3.751M
OpenRouterRouter$1.48$4.431M
Deep InfraHost$2.50$7.50256K
GMI CloudHost$2.50$7.501M
Together AIHost$2.50$7.501M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.16 counted60.565 / 218
CodingCode writing and repair.11 counted59.872 / 209
ReasoningLogic problems and puzzles.7 counted60.464 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.9 counted58.673 / 268
MultilingualTasks in many languages.2 counted57.21 / 9
InstructionTasks with strict prompt rules.3 counted64.48 / 63
MathMath problems.6 counted62.140 / 168

Results

54 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Massive Multitask Language Understanding ProfessionalKnowledge89.6%1 / 75
Artificial Analysis IFBenchInstruction80.5%4 / 146
Harvard-MIT Mathematics Tournament February 2026Math97.1%1 / 27
MMLU-ProXMultilingual87.0%1 / 17
VITA-BenchAgentic47.9%1 / 13
MMLU ProKnowledge89.3% ±0.312 / 135
Artificial Analysis GPQA DiamondKnowledge92.3%19 / 199
Gert Labs Composite Game BenchmarkAgentic64.3%5 / 52
QwenClawBenchAgentic64.3%1 / 10
LiveCodeBench v6Coding91.6%6 / 59
MMLU-ReduxKnowledge95.0%2 / 19
Graduate-Level Google-Proof Q&AKnowledge92.4%16 / 141
Instruction-Following EvalInstruction94.3%8 / 70
Scientific Code BenchmarkCoding53.5%4 / 35
Berkeley Function Calling Leaderboard v4Agentic75.0%3 / 26
Terminal-Bench 2.0 · BenchLMAgentic69.7%5 / 41
MMMLUKnowledge90.3%2 / 15
LiveCodeBenchCoding87.1% ±1.019 / 142
Terminal-Bench 2.0 · Vals AIAgentic59.2% ±1.49 / 66
Software Engineering Benchmark VerifiedCoding80.4%18 / 114
τ²-Bench Tool-Agent-User EvaluationAgentic94.7%24 / 149
GPQA diamond · Epoch AIKnowledge90.9% ±2.035 / 217
IOI v1Coding46.8% ±10.311 / 61
SuperGPQA: Scaling LLM Evaluation Across 285 Graduate DisciplinesKnowledge73.6%4 / 22
GPQA Diamond · Vals AIKnowledge90.2% ±1.525 / 135
SWE-Bench verified · Epoch AICoding77.3% ±1.96 / 32
OTIS Mock AIME 2024-2025Math95.6% ±3.140 / 200
IMOAnswerBenchMath90.0%2 / 10
Artificial Analysis Humanity's Last ExamKnowledge40.5%46 / 222
LiveBench Instruction FollowingInstruction74.0%14 / 66
GPQA Diamond · BenchLMKnowledge92.4%18 / 84
INCLUDEMultilingual86.2%3 / 12
Critical Physics TasksReasoning13.4%54 / 214
SimpleQA VerifiedKnowledge55.8% ±1.624 / 91
Artificial Analysis Long Context ReasoningReasoning79.0%58 / 217
MRCRv2Reasoning90.4%4 / 14
Instruction Following BenchmarkInstruction79.1%19 / 57
MCP AtlasAgentic76.4%20 / 52
SWE-bench ProCoding60.6%32 / 83
Terminal-Bench HardAgentic50.8%9 / 23
Claw-EvalAgentic65.2%17 / 42
FrontierMath-Tier-4-v2-PrivateMath34.1% ±7.530 / 72
Humanity's Last ExamKnowledge41.4%33 / 77
FrontierMath-Tiers-1-3-v2-PrivateMath64.6% ±2.839 / 90
Mystery Game PuzzlesReasoning26.0% ±4.435 / 80
GDPval-AA normalizedAgentic31.4%59 / 133
Artificial Analysis Omniscience AccuracyKnowledge31.1%76 / 164
LiveBench LanguageKnowledge79.7%31 / 66
NL2RepoCoding47.2%17 / 36
Artificial Analysis SciCodeCoding49.5%54 / 114
Terminal-Bench 2.1 · Vals AIAgentic61.0% ±0.439 / 75
LiveBench ReasoningReasoning83.3%46 / 80
Vibe Code Bench v1.1Coding47.7% ±4.671 / 123
Chess PuzzlesReasoning27.0% ±4.552 / 90
Humanity's Last Exam with toolsAgentic53.5%22 / 36
LiveBench Data AnalysisReasoning71.8%50 / 80
Artificial Analysis ITBench-AAAgentic42.5%15 / 24
Agent Arena command recoveryAgentic−1.637 / 50
LiveBench CodingCoding74.2%50 / 66
OpenHarmony Bench v1.0Coding53.4%10 / 13
LiveBench MathematicsMath85.2%51 / 66
Code MigrationCoding13.1% ±2.960 / 77
SWE-benchCoding68.8% ±2.168 / 87
LiveBench Agentic CodingAgentic43.6%55 / 66
Agent Arena task outcomeAgentic−10.545 / 54
EBR-benchReasoning9.5%26 / 31
Agent Arena steerabilityAgentic−12.848 / 54
TaxEval v2Finance75.3% ±0.819 / 143
LegalBenchLegal84.9% ±0.428 / 149
CorpFin v2Finance63.7% ±0.940 / 131
EMBFinance57.0% ±2.942 / 75
MedScribeHealthcare79.4% ±1.963 / 109
Finance Agent (v2)Finance47.8% ±0.752 / 79
MedCodeHealthcare38.8% ±2.270 / 106
Legal Research BenchLegal25.5% ±3.063 / 90
Tax Agent BenchFinance15.8% ±2.659 / 82
MAXIFEMultilingual89.2%—
PolyMathMultilingual86.5%—
NOVA-63Multilingual59.0%—
Artificial Analysis EnterpriseOps-GymAgentic45.0%—
ApexMath44.5%—
ResearchClawBenchAgentic18.7%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYVals AINo licence statedLiveBenchNo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index61.6 → 60.5
23 Oct 202612 Nov 2026
Output / 1M$4.43 → $4.43
23 Oct 202612 Nov 2026

Questions

How much does Qwen3.7 Max cost?
$1.48 per 1M input tokens, $4.43 per 1M output tokens.
How good is Qwen3.7 Max?
Its index is 60.5 out of 100. That is place 69 of 283 ranked models.
What is Qwen3.7 Max best at?
Multilingual: 57.2 out of 100, place 1 of 9.
How fast is Qwen3.7 Max?
68 output tokens per second, 1.3 s to the first token.
What is the context window of Qwen3.7 Max?
1M tokens.
Is Qwen3.7 Max open source?
No. Its weights are not published.
When was Qwen3.7 Max released?
May 2026.

Badge

Qwen3.7 Max on Inference360: place 69 of 283, index 60.5
HTML
<a href="https://inference360.ai/models/qwen3-7-max"><img src="https://inference360.ai/badges/qwen3-7-max.svg" alt="Qwen3.7 Max on Inference360: place 69 of 283, index 60.5" height="20"></a>
Markdown
[![Qwen3.7 Max on Inference360: place 69 of 283, index 60.5](https://inference360.ai/badges/qwen3-7-max.svg)](https://inference360.ai/models/qwen3-7-max)

Compare with

Sage-A1.560.5GLM-5.260.6DeepSeek V4 Flash Vision Exp60.4Mistral Large 460.7Qwen3.7 Plus57.7

Lower price

dots3-note Preview63.4 · FreeGLM-5.3-Flash63.1 · $0.078/1MLing 3.1 Flash67.2 · $0.135/1M

More from Alibaba

All models
Qwen3.7 Plus57.7Qwen3.6 FlashUnrankedQwen3.6-27B52.1Qwen 3.6 Max (preview)57.2Qwen3.6-35B-A3B50.6