Skip to content

Qwen3.6-27B

AlibabaOpen weightsReasoningProvisionalReleased Apr 21, 2026Input: text, image, video
Compare with Quasar 438BSource scores ↗OpenRouter ↗
Overall score
48.6
Rank 108 of 222 · 34.2–62.9
Context window
262K
262,000 tokens
Price per 1M tokens
$0.25 in · $2.00 out
Source: OpenRouter
Blended price
$0.69
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.

  1. 1Instruction following#5582.4
  2. 2Reasoning78.1
  3. 3Multimodal#3967.3
  4. 4Math#865.9
  5. 5Knowledge#9254.3
  6. 6Agentic & tool use54.2
  7. 7Coding#8047.9
Higher is better. #N is the model's rank in that category, where it is eligible to be ranked.

Benchmark results (38)

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

Multimodal

MMMU-Pro75.8#30 of 48
CharXiv78.4#31 of 42
MMMU82.9#7 of 21
SimpleVQA56.1#15 of 17
RealWorldQA84.1#7 of 16
ERQA62.5#12 of 16
VideoMMMU84.4#9 of 12
V*94.7#5 of 12
Video-MME (with subtitle)87.7#4 of 11
MLVU (M-Avg)86.6#4 of 9
CountBench97.8#1 of 8
RefCOCO (avg)92.5#1 of 8
CC-OCR81.2#3 of 7
DynaMath85.6#3 of 6
MStar81.4#1 of 1

Agentic & tool use

Gert Labs54.84#17 of 52
Terminal-Bench 2.059.3#18 of 39
Claw-Eval72.4#7 of 36
QwenClawBench53.4#7 of 10
AndroidWorld70.3#7 of 8
QwenWebBench1487#4 of 5

Knowledge

GPQA87.8#36 of 117
HLE24.0#55 of 69
MMLU-Pro86.2#12 of 67
SuperGPQA66.0#13 of 20
MMLU-Redux93.5#7 of 15
C-Eval91.4#5 of 10

Math

FrontierMath v2 (Tiers 1-3)35.088#49 of 64
AIME2694.1#17 of 29
HMMT Feb 202684.3#16 of 24
HMMT Feb 202593.8#7 of 13
HMMT Nov 202590.7#8 of 12
MMAnswerBench80.8#9 of 10

Coding

SWE-bench Verified77.2#33 of 98
SWE-bench Pro53.5#52 of 69
LiveCodeBench v683.9#20 of 46
NL2Repo36.2#26 of 31

Multilingual

SWE Multilingual71.3#34 of 46

Pricing by source

USD per 1M tokens.

SourceInputOutputCached in
List price (OpenRouter)$0.25$2.00—
OpenRouter qwen/qwen3.6-27b$0.25$2.00$0.025

Self-hosted open-weight model. Qwen published both the base weights and an FP8 checkpoint on April 21, 2026; public hosted pricing varies by provider.

Open weights

From the Hugging Face repository.

Parameters
28B
Licence
apache-2.0
Downloads (30 days)
2.3M

Speed

Measured by us through OpenRouter (median of trials).

Not measured yet.

Related models

Same family

Similar score