Inference
EN
Alibaba

Qwen3.6-35B-A3B

Qwen3.6-35B-A3B is a reasoning model from Alibaba. 39 benchmarks count toward its index, in 7 categories.

  • Open weights
  • Reasoning
  • Input: text · image · video
  • Released 2026-04
RankedEnough results for an index.
IndexAll categories, out of 100.50.6±3.1Median 51.0
CoverageIndex weight that has results.95%
SpeedOutput tokens per second.65/sMedian 67/s
First tokenSeconds to the first output token.0.6 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.10Median $0.60
Output / 1MUS dollars per 1M output tokens.$0.90Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.—Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
qwen/qwen3.6-35b-a3b
Max outputMaximum output tokens per request.
236K
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.
$0.025
WeightsLink to the published weights.
Hugging Face (opens in a new tab) ↗
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
6 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
CoreWeaveHost$0.10$0.89262K
OpenRouterRouter$0.10$0.90262K
Deep InfraHost$0.10$0.95262K
Novita AIHost$0.248$1.49262K
ScalewayCloud$0.28$1.68128K
SiliconFlowHost$0.24$1.80262K

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

Run locally

35.95B parameters
16-bit
76.5 GB
8-bit
41.1 GB
4-bit
23.9 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 24 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.11 counted52.7114 / 218
CodingCode writing and repair.5 counted49.6140 / 209
ReasoningLogic problems and puzzles.3 counted49.9117 / 227
MultimodalTasks with images and text.8 counted51.565 / 84
KnowledgeFacts and expert knowledge.5 counted49.8137 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted49.2—
MathMath problems.6 counted49.092 / 168

Results

39 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
τ²-Bench Tool-Agent-User EvaluationAgentic95.3%21 / 149
Massive Multitask Language Understanding ProfessionalKnowledge85.2%18 / 75
Claw-EvalAgentic68.7%12 / 42
RefCOCO averageMultimodal92.0%3 / 10
Massive Multi-discipline Multimodal UnderstandingMultimodal81.7%9 / 28
GPQA diamond · Epoch AIKnowledge84.8% ±2.672 / 217
RealWorldQAMultimodal85.3%6 / 18
Graduate-Level Google-Proof Q&AKnowledge86.0%48 / 141
τ³-Bench Tool-Agent-User EvaluationAgentic67.2%11 / 32
Artificial Analysis IFBenchInstruction64.4%57 / 146
Artificial Analysis GPQA DiamondKnowledge84.1%85 / 199
Software Engineering Benchmark VerifiedCoding73.4%53 / 114
Artificial Analysis MMMU-ProMultimodal75.0%53 / 113
Artificial Analysis Humanity's Last ExamKnowledge22.2%106 / 222
Artificial Analysis Long Context ReasoningReasoning71.7%106 / 217
LiveCodeBench v6Coding80.4%29 / 59
C-EvalKnowledge90.0%6 / 12
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)MultimodalAbove fitFar above this model's index.88.1%13 / 26
OTIS Mock AIME 2024-2025Math68.9% ±7.0102 / 200
Mystery Game PuzzlesReasoning22.0% ±4.241 / 80
Gert Labs Composite Game BenchmarkAgentic42.6%32 / 52
VITA-BenchAgentic35.6%8 / 13
GDPval-AA normalizedAgentic19.6%82 / 133
Critical Physics TasksReasoning0.3%137 / 214
Terminal-Bench 2.0 · BenchLMAgentic51.5%27 / 41
Massive Multi-discipline Multimodal Understanding ProMultimodal75.3%35 / 53
Harvard-MIT Mathematics Tournament February 2025Math90.7%10 / 15
MCP AtlasAgentic62.8%35 / 52
AIME 2026Math92.7%23 / 34
Harvard-MIT Mathematics Tournament February 2026Math83.6%19 / 27
SimpleVQAMultimodal58.9%13 / 18
Harvard-MIT Mathematics Tournament November 2025Math89.1%9 / 12
SuperGPQA: Scaling LLM Evaluation Across 285 Graduate DisciplinesKnowledge64.7%17 / 22
CharXiv ReasoningMultimodal78.0%39 / 49
QwenClawBenchAgentic52.6%8 / 10
Humanity's Last ExamKnowledge21.4%62 / 77
FrontierMath-Tiers-1-3-v2-PrivateMath20.4% ±2.477 / 90
Artificial Analysis SciCodeCoding36.6%98 / 114
SWE-bench ProCoding49.5%72 / 83
WideResearchAgentic60.1%15 / 16
NL2RepoCoding29.4%35 / 36
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal37.4%39 / 39
ToolathlonAgentic26.9%22 / 22
AI2D test splitMultimodal92.7%—
OmniDocBench 1.5Multimodal89.9%—
Video-MME with subtitleMultimodal86.6%—
MLVU mean averageMultimodal86.2%—
VideoMMMUMultimodal83.7%—
Video-MME without subtitleMultimodal82.5%—
CC-OCRMultimodal81.9%—
MMAnswerBenchMath78.9%—
ODINW13Multimodal50.8%—
DeepPlanningAgentic25.9%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYmodels.devMIT

History

5 releases
Index51.1 → 50.6
23 Oct 202612 Nov 2026
Output / 1M$1.00 → $0.90
23 Oct 202612 Nov 2026

Questions

How much does Qwen3.6-35B-A3B cost?
$0.10 per 1M input tokens, $0.90 per 1M output tokens.
How good is Qwen3.6-35B-A3B?
Its index is 50.6 out of 100. That is place 147 of 283 ranked models.
What is Qwen3.6-35B-A3B best at?
Knowledge: 49.8 out of 100, place 137 of 268.
How fast is Qwen3.6-35B-A3B?
65 output tokens per second, 0.6 s to the first token.
What is the context window of Qwen3.6-35B-A3B?
262K tokens.
Is Qwen3.6-35B-A3B open source?
Yes. Its weights are published.
When was Qwen3.6-35B-A3B released?
April 2026.

Badge

Qwen3.6-35B-A3B on Inference360: place 147 of 283, index 50.6
HTML
<a href="https://inference360.ai/models/qwen3-6-35b-a3b"><img src="https://inference360.ai/badges/qwen3-6-35b-a3b.svg" alt="Qwen3.6-35B-A3B on Inference360: place 147 of 283, index 50.6" height="20"></a>
Markdown
[![Qwen3.6-35B-A3B on Inference360: place 147 of 283, index 50.6](https://inference360.ai/badges/qwen3-6-35b-a3b.svg)](https://inference360.ai/models/qwen3-6-35b-a3b)

Compare with

Nemotron 3 Ultra50.8Nemotron 3.5 Super VL 120B A12B NVFP450.8Mistral Medium 3.5 128B50.3Qwen3.6 Plus54.6Qwen 3.6 Max (preview)57.2

Lower price

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

More from Alibaba

All models
Qwen 3.6 Max (preview)57.2Qwen3.6-27B52.1Qwen3.6 FlashUnrankedQwen3.6 Plus54.6Qwen3.7 Max60.5