Inference
EN

Qwen3.5-35B-A3B vs Qwen3.6 Plus

Change models
Qwen3.5-35B-A3BAlibaba
Qwen3.6 PlusAlibaba
IndexAll categories, out of 100.
49.4 ±3.2
54.6 ±2.1 best
GapIndex points below the top.
−5.2
Top
CoverageIndex weight that has results.
100%
100%
SpeedOutput tokens per second.
43/s
44/s best
EloLMArena rating and rank.
1395
1436 best
ReleasedMonth of release.
2026-03
2026-04
Input / 1MUS dollars per 1M input tokens.
$0.14 best
$0.325
Output / 1MUS dollars per 1M output tokens.
$1.00 best
$1.95
1K chats2K tokens in, 500 out, each.
$0.78 best
$1.63
1K documents20K tokens in, 1K out, each.
$3.80 best
$8.45
1K agent tasks200K tokens in, 20K out, each.
$48.00 best
$104.00
ContextMaximum tokens in one request.
262K
1M best
AgenticMulti-step tasks with tools.
47.3
55.0 best
CodingCode writing and repair.
49.0
56.2 best
ReasoningLogic problems and puzzles.
48.5
51.2 best
MultimodalTasks with images and text.
55.1
56.1 best
KnowledgeFacts and expert knowledge.
47.4
54.1 best
MultilingualTasks in many languages.
50.4
54.1 best
InstructionTasks with strict prompt rules.
52.5 best
52.4
MathMath problems.
49.6
54.9 best

Benchmarks

25 shared
Only Qwen3.5-35B-A3B: 6Only Qwen3.6 Plus: 66
Qwen3.5-35B-A3B
Qwen3.6 Plus
Artificial Analysis Humanity's Last ExamKnowledge
21.0%111 / 222
27.8% best92 / 222
Artificial Analysis Long Context ReasoningReasoning
72.0%105 / 217
78.3% best63 / 217
GPQA diamond · Epoch AIKnowledge
81.2%86 / 217
88.4% best54 / 217
Critical Physics TasksReasoning
0.9%124 / 214
2.9% best98 / 214
OTIS Mock AIME 2024-2025Math
70.0%100 / 200
93.3% best51 / 200
Artificial Analysis GPQA DiamondKnowledge
84.5%80 / 199
88.2% best52 / 199
Artificial Analysis Omniscience AccuracyKnowledge
20.1%130 / 164
26.4% best95 / 164
τ²-Bench Tool-Agent-User EvaluationAgentic
89.2%44 / 149
97.7% best9 / 149
Artificial Analysis IFBenchInstruction
72.5%35 / 146
75.2% best25 / 146
Graduate-Level Google-Proof Q&AKnowledge
84.2%55 / 141
90.4% best27 / 141
Software Engineering Benchmark VerifiedCoding
69.2%76 / 114
78.8% best26 / 114
Artificial Analysis MMMU-ProMultimodal
72.7%68 / 113
78.0% best39 / 113
Massive Multitask Language Understanding ProfessionalKnowledge
85.3%17 / 75
88.5% best3 / 75
Instruction-Following EvalInstruction
91.9%21 / 70
94.3% best8 / 70
LiveCodeBench v6Coding
74.6%35 / 59
87.1% best17 / 59
Massive Multi-discipline Multimodal Understanding ProMultimodal
75.1%37 / 53
78.8% best21 / 53
Gert Labs Composite Game BenchmarkAgentic
29.0%51 / 52
50.6% best21 / 52
CharXiv ReasoningMultimodal
77.5%40 / 49
81.5% best27 / 49
Terminal-Bench 2.0 · BenchLMAgentic
40.5%38 / 41
61.6% best16 / 41
Massive Multi-discipline Multimodal UnderstandingMultimodal
81.4%10 / 28
86.0% best3 / 28