← All modelsopen_source_balanced
Qwen 3 235B
Alibaba (via hosted providers)
Capabilities
ToolsCodeExtended ThinkingStructured OutputMultilingual
Strengths
- Hybrid thinking (toggle reasoning on/off)
- Excellent multilingual
- MoE (22B active)
- Strong benchmark scores
- #1 open model on text (lmarena.ai)
- Apache 2.0
Weaknesses
- No vision (use Qwen3-VL variant for vision)
- Self-hosting needs significant GPU
Pricing
Input / 1M tokens
$0.4
Output / 1M tokens
$1.2
Context window
128k
Transparency
Open weights
10.0 / 10
Open training data
3.0 / 10
Open methodology
7.0 / 10
Licence openness
10.0 / 10
Provider disclosure
2.0 / 10
FMTI company score
300.0 / 10
Composite:6.4 / 10
Apache 2.0, open weights. Alibaba FMTI 30.
Sustainability
Inference energy
5.0 / 10
Training footprint
N/A
Provider infrastructure
N/A
Composite:5.0 / 10
MoE efficient. Depends on hosting.
Task Fitness
Code
73%
Vision
0%
Analyse
73%
Extract
70%
Generate
68%
Summarise
70%
Translate
76%
Conversation
68%