Qwen3
Alibaba·
LLMsopen-weightcloud + localMulti-GPU Self-Host
Overview
235B MoE flagship with hybrid thinking modes.
Capabilities and innovations
235B MoEHybrid thinking119 languages128k contextThinking/Non-thinking ModesMoE ArchitectureAgentic Coding
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 68.9% | unsourced |
| SWE-bench Verified | 42% | unsourced |
| MMLU | 89.5% | unsourced |
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Pro: benchmark did not exist at release
- MMLU-Pro: not reported by the vendor
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
Architecture and hardware
- Parameters
- 235B total, 22B active per token (MoE)
- Estimated VRAM at Q4
- ~136 GB, Multi-GPU Self-Host class
Links
More from Alibaba
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.