Inference
EN

Qwen3.6-27B vs Qwen3.7 Max

Change models
Qwen3.6-27BAlibaba
Qwen3.7 MaxAlibaba
IndexAll categories, out of 100.
52.1 ±2.4
60.5 ±3.2 best
GapIndex points below the top.
−8.4
Top
CoverageIndex weight that has results.
100% best
90%
SpeedOutput tokens per second.
60/s
68/s best
EloLMArena rating and rank.
—
1475
ReleasedMonth of release.
2026-04
2026-05
Input / 1MUS dollars per 1M input tokens.
$0.32 best
$1.48
Output / 1MUS dollars per 1M output tokens.
$3.20 best
$4.43
1K chats2K tokens in, 500 out, each.
$2.24 best
$5.16
1K documents20K tokens in, 1K out, each.
$9.60 best
$33.93
1K agent tasks200K tokens in, 20K out, each.
$128.00 best
$383.50
ContextMaximum tokens in one request.
262K
1M best
AgenticMulti-step tasks with tools.
56.7
60.5 best
CodingCode writing and repair.
52.8
59.8 best
ReasoningLogic problems and puzzles.
49.0
60.4 best
MultimodalTasks with images and text.
53.4
—
KnowledgeFacts and expert knowledge.
48.2
58.6 best
MultilingualTasks in many languages.
60.9 best
57.2
InstructionTasks with strict prompt rules.
43.2
64.4 best
MathMath problems.
51.0
62.1 best

Benchmarks

39 shared
Only Qwen3.6-27B: 25Only Qwen3.7 Max: 43
Qwen3.6-27B
Qwen3.7 Max
Artificial Analysis Humanity's Last ExamKnowledge
23.1%104 / 222
40.5% best46 / 222
Artificial Analysis Long Context ReasoningReasoning
77.3%77 / 217
79.0% best58 / 217
GPQA diamond · Epoch AIKnowledge
84.8%72 / 217
90.9% best35 / 217
Critical Physics TasksReasoning
1.1%118 / 214
13.4% best54 / 214
OTIS Mock AIME 2024-2025Math
66.7%106 / 200
95.6% best40 / 200
Artificial Analysis GPQA DiamondKnowledge
84.2%83 / 199
92.3% best19 / 199
Artificial Analysis Omniscience AccuracyKnowledge
19.6%134 / 164
31.1% best76 / 164
τ²-Bench Tool-Agent-User EvaluationAgentic
94.2%26 / 149
94.7% best24 / 149
Artificial Analysis IFBenchInstruction
67.6%53 / 146
80.5% best4 / 146
TaxEval v2Finance
71.3%79 / 143
75.3% best19 / 143
Graduate-Level Google-Proof Q&AKnowledge
87.8%40 / 141
92.4% best16 / 141
GDPval-AA normalizedAgentic
24.3%75 / 133
31.4% best59 / 133
CorpFin v2Finance
62.3%47 / 131
63.7% best40 / 131
Vibe Code Bench v1.1Coding
11.9%103 / 123
47.7% best71 / 123
Artificial Analysis SciCodeCoding
42.8%79 / 114
49.5% best54 / 114
Software Engineering Benchmark VerifiedCoding
77.2%36 / 114
80.4% best18 / 114
FrontierMath-Tiers-1-3-v2-PrivateMath
34.0%65 / 90
64.6% best39 / 90
SWE-benchCoding
70.0% best61 / 87
68.8%68 / 87
SWE-bench ProCoding
53.5%60 / 83
60.6% best32 / 83
LiveBench Data AnalysisReasoning
70.4%55 / 80
71.8% best50 / 80