Pyyan / Compare / DeepSeek V4-Pro vs Kimi K3 vs GLM-5.2 vs Qwen3-235B-A22B vs Llama 5
DeepSeek V4-Pro vs Kimi K3 vs GLM-5.2 vs Qwen3-235B-A22B vs Llama 5
5 of 5Open-Weight Models · verified 13 Aug 2026
×DeepSeek V4-ProDeepSeekcurrent
×Kimi K3Moonshot AIcurrent
×Qwen3-235B-A22BAlibabacurrent
×Llama 5Metacurrent
| Specification | DeepSeek V4-Pro | Kimi K3 | GLM-5.2 | Qwen3-235B-A22B | Llama 5 |
|---|---|---|---|---|---|
| Summary | Best all-round open model of 2026. | Largest open-weight model shipped. | Strongest open reasoning and mathematics. | Cheapest capable model, most permissive licence. | Open-weight flagship with recursive self-improvement. |
| Licence | MIT | Open weights | Open weights | Apache-2.0 | Llama licence |
| Parameters | — | 2.8T | — | 235B MoE | 600B+ |
| Released | — | 17 Jul 2026 | — | — | — |
| Category | Open-Weight Models | Open-Weight Models | Open-Weight Models | Open-Weight Models | Open-Weight Models |
| Official | DeepSeek ↗ | Moonshot AI ↗ | Zhipu AI ↗ | Alibaba ↗ | Meta ↗ |
Highlighted rows are where these differ.
DeepSeek V4-Pro
- Tops open leaderboards on agentic coding and reasoning
- MIT licence — no conditions
- Pioneered aggressive cache-hit pricing at ~$0.07/M
Best for everything open.
Qwen3-235B-A22B
- No licence conditions whatsoever
- 22B active of 235B total
Best for commercial self-hosting.
Llama 5
- Refines its own reasoning and generates synthetic training data
Best for open frontier.