Qwen: Qwen3.8 Flash

Alibaba

budget

Capabilities

VisionToolsStructured OutputExtended ThinkingLong ContextCode

Strengths

Exceptional speed at 56 tok/s for a reasoning-capable model
1M token context window enables long-document and video analysis
Strong coding and agentic task performance (0.86–0.93)
Multimodal (text, image, video) with extended thinking
Very affordable pricing (<$1/M output)

Weaknesses

Closed-source; weights and training data unavailable
Limited provider transparency on methodology and governance
May struggle with nuanced reasoning vs. flagship alternatives
Minimal sustainability or energy efficiency disclosure

Pricing

Input / 1M tokens

$0.15

Output / 1M tokens

$0.47

Context window

1,000k

Transparency

Open weights

0.0 / 10

Open training data

0.0 / 10

Open methodology

4.0 / 10

Licence openness

1.0 / 10

Provider disclosure

4.5 / 10

FMTI company score

N/A

Composite:6.5 / 10

Closed-source model with minimal weight/training data disclosure. Alibaba publishes some technical details but methodology remains proprietary. Licensing is restricted.

Sustainability

Inference energy

N/A

Training footprint

N/A

Provider infrastructure

5.0 / 10

Composite:5.0 / 10

Alibaba's sustainability commitments are opaque. No public data on inference energy or training carbon footprint. Provider infrastructure score reflects lack of disclosed green commitments.

Task Fitness

question answering

67%

code

86%

mathematics

50%

business communication

50%

vision

78%

analysis

67%

extraction

80%

long-form generation

70%

research

50%

reasoning

80%

summarisation

82%

translation

72%

conversation

75%