Inference
EN
Alibaba

Qwen3.5-35B-A3B

Qwen3.5-35B-A3B is a reasoning model from Alibaba. 25 benchmarks count toward its index, in 8 categories.

  • Open weights
  • Reasoning
  • Input: text · image · video
  • Released 2026-03
RankedEnough results for an index.
IndexAll categories, out of 100.49.4±3.2Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.43/sMedian 67/s
First tokenSeconds to the first output token.0.5 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.14Median $0.60
Output / 1MUS dollars per 1M output tokens.$1.00Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.1395±4 #166Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
qwen/qwen3.5-35b-a3b
Max outputMaximum output tokens per request.
82K
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.
$0.05
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.
4 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Deep InfraHost$0.14$1.00262K
OpenRouterRouter$0.14$1.00262K
SiliconFlowHost$0.24$1.80262K
Novita AIHost$0.25$2.00262K

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.5 counted47.3152 / 218
CodingCode writing and repair.2 counted49.0144 / 209
ReasoningLogic problems and puzzles.3 counted48.5127 / 227
MultimodalTasks with images and text.6 counted55.153 / 84
KnowledgeFacts and expert knowledge.5 counted47.4154 / 268
MultilingualTasks in many languages.1 counted50.4—
InstructionTasks with strict prompt rules.2 counted52.534 / 63
MathMath problems.1 counted49.6—

Results

25 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Massive Multitask Language Understanding ProfessionalKnowledge85.3%17 / 75
Artificial Analysis IFBenchInstruction72.5%35 / 146
τ²-Bench Tool-Agent-User EvaluationAgentic89.2%44 / 149
Instruction-Following EvalInstruction91.9%21 / 70
Massive Multi-discipline Multimodal UnderstandingMultimodal81.4%10 / 28
Graduate-Level Google-Proof Q&AKnowledge84.2%55 / 141
GPQA diamond · Epoch AIKnowledge81.2% ±2.386 / 217
Artificial Analysis GPQA DiamondKnowledge84.5%80 / 199
Artificial Analysis Long Context ReasoningReasoning72.0%105 / 217
OTIS Mock AIME 2024-2025Math70.0% ±6.0100 / 200
Artificial Analysis Humanity's Last ExamKnowledge21.0%111 / 222
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)MultimodalAbove fitFar above this model's index.88.0%14 / 26
V*Multimodal92.7%8 / 14
Critical Physics TasksReasoning0.9%124 / 214
LiveCodeBench v6Coding74.6%35 / 59
Artificial Analysis MMMU-ProMultimodal72.7%68 / 113
MMLU-ProXMultilingual81.0%11 / 17
Software Engineering Benchmark VerifiedCoding69.2%76 / 114
Massive Multi-discipline Multimodal Understanding ProMultimodal75.1%37 / 53
LongBench v2Reasoning59.0%18 / 25
MathVisionMultimodal83.9%18 / 24
Artificial Analysis Omniscience AccuracyKnowledge20.1%130 / 164
BrowseCompAgentic61.0%51 / 63
CharXiv ReasoningMultimodal77.5%40 / 49
SuperGPQA: Scaling LLM Evaluation Across 285 Graduate DisciplinesKnowledge63.4%18 / 22
Terminal-Bench 2.0 · BenchLMAgentic40.5%38 / 41
OSWorld-VerifiedAgentic54.5%35 / 37
Gert Labs Composite Game BenchmarkAgentic29.0%51 / 52
Artificial Analysis Global-MMLU-LiteMultilingual86.4%—
Multimodal Multi-disciplinary Video UnderstandingMultimodal72.3%—
SWE-RebenchCoding53.7%—

Sources

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

History

5 releases
Index50.3 → 49.4
23 Oct 202612 Nov 2026
Output / 1M$1.00 → $1.00
23 Oct 202612 Nov 2026

Questions

How much does Qwen3.5-35B-A3B cost?
$0.14 per 1M input tokens, $1.00 per 1M output tokens.
How good is Qwen3.5-35B-A3B?
Its index is 49.4 out of 100. That is place 153 of 283 ranked models.
What is Qwen3.5-35B-A3B best at?
Instruction: 52.5 out of 100, place 34 of 63.
How fast is Qwen3.5-35B-A3B?
43 output tokens per second, 0.5 s to the first token.
What is the context window of Qwen3.5-35B-A3B?
262K tokens.
Is Qwen3.5-35B-A3B open source?
Yes. Its weights are published.
When was Qwen3.5-35B-A3B released?
March 2026.

Badge

Qwen3.5-35B-A3B on Inference360: place 153 of 283, index 49.4
HTML
<a href="https://inference360.ai/models/qwen3-5-35b-a3b"><img src="https://inference360.ai/badges/qwen3-5-35b-a3b.svg" alt="Qwen3.5-35B-A3B on Inference360: place 153 of 283, index 49.4" height="20"></a>
Markdown
[![Qwen3.5-35B-A3B on Inference360: place 153 of 283, index 49.4](https://inference360.ai/badges/qwen3-5-35b-a3b.svg)](https://inference360.ai/models/qwen3-5-35b-a3b)

Compare with

Grok 450.0Qwen3.5-27B51.3Qwen3.6 Plus54.6dots3-note Preview63.4

Lower price

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

More from Alibaba

All models
Qwen3.5-122B-A10B51.9Qwen3.5-27B51.3Qwen3.5 FlashUnrankedQwen3.5 PlusUnrankedQwen3.5 9BUnranked