Qwen3.6-35B-A3B

AlibabaOpen weightsReasoningProvisionalReleased Apr 15, 2026Input: text, image, video

Compare with Grok 4 Fast (Reasoning)Source scores ↗OpenRouter ↗

Overall score

43.5

Rank 128 of 222 · 29.1–57.9

Context window

262K

262,000 tokens

Price per 1M tokens

$0.10 in · $0.95 out

Source: OpenRouter

Blended price

$0.31

3:1 input:output tokens

Output speed

—

Not measured yet

Category scores

Scores per capability area with the model's rank among ranked models in that category.

Higher is better. #N is the model's rank in that category, where it is eligible to be ranked.

Benchmark results (43)

As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.

Multimodal

MMMU-Pro	75.3	#31 of 48
CharXiv	78.0	#32 of 42
ParseBench	44.10	#34 of 37
ExtractBench (LlamaIndex)(lower is better)	88.11	#14 of 24
MMMU	81.7	#9 of 21
SimpleVQA	58.9	#13 of 17
RealWorldQA	85.3	#6 of 16
VideoMMMU	83.7	#12 of 12
OmniDocBench 1.5	89.9	#6 of 11
Video-MME (with subtitle)	86.6	#8 of 11
MLVU (M-Avg)	86.2	#5 of 9
RefCOCO (avg)	92.0	#3 of 8
CC-OCR	81.9	#2 of 7
AI2D_TEST	92.7	#3 of 7
Video-MME (w/o subtitle)	82.5	#5 of 7
ODINW13	50.8	#2 of 4

Agentic & tool use

Gert Labs	42.65	#32 of 52
MCP Atlas	62.8	#28 of 42
Terminal-Bench 2.0	51.5	#27 of 39
Claw-Eval	68.7	#9 of 36
τ³-bench results	67.2	#10 of 23
Toolathlon	26.9	#22 of 22
WideResearch	60.1	#15 of 16
VITA-Bench	35.6	#6 of 11
QwenClawBench	52.6	#8 of 10
DeepPlanning	25.9	#5 of 7
QwenWebBench	1397	#5 of 5

Math

FrontierMath v2 (Tiers 1-3)	20.351	#58 of 64
AIME26	92.7	#20 of 29
HMMT Feb 2026	83.6	#17 of 24
HMMT Feb 2025	90.7	#10 of 13
HMMT Nov 2025	89.1	#9 of 12
MMAnswerBench	78.9	#10 of 10

Knowledge

GPQA	86.0	#44 of 117
HLE	21.4	#57 of 69
MMLU-Pro	85.2	#18 of 67
SuperGPQA	64.7	#15 of 20
C-Eval	90.0	#6 of 10

Coding

SWE-bench Verified	73.4	#47 of 98
SWE-bench Pro	49.5	#62 of 69
LiveCodeBench v6	80.4	#25 of 46
NL2Repo	29.4	#31 of 31

Multilingual

SWE Multilingual

67.2

#38 of 46

Pricing by source

USD per 1M tokens.

Source	Input	Output	Cached in
List price (OpenRouter)	$0.10	$0.95	—
OpenRouter qwen/qwen3.6-35b-a3b	$0.10	$0.95	$0.050

Price history

Oct 17, 2026: $0.15 / $1.00
Oct 20, 2026: $0.10 / $0.95

Open weights

From the Hugging Face repository.

Parameters: 36B
Licence: apache-2.0
Downloads (30 days): 3.6M
Repository: Qwen/Qwen3.6-35B-A3B ↗

Speed

Measured by us through OpenRouter (median of trials).

Not measured yet.

Related models

Same family

Qwen3.6 Plus53.8
Qwen3.5-35B-A3B45.8

Similar score

Grok 4 Fast (Reasoning)43.5 · —
Step 3.7 Flash43.6 · $0.44
Grok 3 Mini44.2 · $0.35
Qwen3.5 Flash42.8 · $0.17
Step 3.5 Flash42.1 · $0.15