← All models

Qwen: Qwen3.6 27B

Alibaba

open_source_balanced

Capabilities

VisionToolsStructured OutputExtended ThinkingLong ContextCode

Strengths

  • Strong long-context handling (262K tokens)
  • Multimodal video+image+text input capability
  • Self-hostable with open weights
  • Extended thinking for complex reasoning
  • Structured output and tool support

Weaknesses

  • Slow inference speed (65 tok/s)
  • Modest reasoning performance (0.63 fitness)
  • Video understanding untested on standard benchmarks
  • Very high output pricing ($3.17/M tokens)

Pricing

Input / 1M tokens

$0.28850000000000003

Output / 1M tokens

$3.17

Context window

262.144k

Transparency

Open weights

10.0 / 10

Open training data

0.0 / 10

Open methodology

7.0 / 10

Licence openness

8.5 / 10

Provider disclosure

6.5 / 10

FMTI company score

N/A

Composite:6.5 / 10

Open-weight model with public weights available. Alibaba publishes moderate technical documentation on architecture and training approaches. Training data composition not fully disclosed. Apache 2.0 licence enables commercial use and self-hosting.

Sustainability

Inference energy

N/A

Training footprint

N/A

Provider infrastructure

5.5 / 10

Composite:5.0 / 10

27B parameter model has moderate inference energy requirements compared to larger models. Alibaba cloud infrastructure sustainability commitments not extensively documented. Self-hosting option available for operators with clean energy sources.

Task Fitness

question answering
55%
code
68%
mathematics
50%
business communication
50%
vision
62%
analysis
58%
extraction
63%
long-form generation
65%
research
50%
reasoning
63%
summarisation
72%
translation
64%
conversation
66%