Best open-weight LLM for knowledge
- Kimi K3 leads
- 68.5
- models ranked
- 18
- downloads of #1, 30 days
- 988.7K
- lead over #2
- +0.9
Knowledge Index · read October 21
- New DeepSeek V4.1 Flash 1017 October 17
- New E Glimpse October 17
- New Mercury 3 October 16
- New R Beam 1.0 501B A23B October 16
As of October 21, 2026, MoonshotAI Kimi K3 leads 18 open-weight models on Knowledge Index with 68.5, ahead of Z.ai GLM 5.3 Flash (67.6) and DeepSeek V4.1 Flash (66.0).
The full best LLM for knowledge ranking combines this benchmark with others.
| Rank | Model | Knowledge Index | Coding evidence | Weights | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | Kimi K3 Moonshot AI | 68.5 | moonshotai/Kimi-K3 | 62.3 | $15.00 | |
| 2 | GLM 5.3 Flash Z.ai | 67.6 | zai-org/GLM-5.3-Flash | 57.7 | $0.50 | |
| 3 | DeepSeek V4.1 Flash DeepSeek | 66.0 | deepseek-ai/DeepSeek-V4.1-Flash | 62.5 | $0.78 | |
| 4 | DeepSeek V4 Pro 0813 DeepSeek | 64.5 | deepseek-ai/DeepSeek-V4-Pro-0813 | 56.1 | $1.98 | |
| 5 | MiniMax M3 MiniMax | 64.3 | MiniMaxAI/Minimax-M3 | 39.2 | $1.20 | |
| 6 | DeepSeek V4.1 Flash 1017 DeepSeek | 62.6 | deepseek-ai/DeepSeek-V4.1-Flash-1017 | 70.3 | $0.60 | |
| 7 | GLM 5.3 Z.ai | 61.3 | zai-org/GLM-5.3 | 70.6 | $4.40 | |
| 8 | DeepSeek V4 Pro 0423 DeepSeek | 60.8 | deepseek-ai/DeepSeek-V4-Pro | 38.7 | $2.716 | |
| 9 | Qwen3.8 27B Qwen | 60.1 | Qwen/Qwen3.8-27B | 74.2 | $2.55 | |
| 10 | T Inkling Thinking Machines | 56.7 | thinkingmachines/Inkling | 31.5 | $4.05 | |
| 11 | Kimi K2.6 Moonshot AI | 53.0 | moonshotai/Kimi-K2.6 | 52.1 | $4.00 | |
| 12 | GLM 5.1 Z.ai | 51.7 | zai-org/GLM-5.1 | 49.2 | $4.40 | |
| 13 | GLM 5.2 Z.ai | 51.6 | zai-org/GLM-5.2 | 61.1 | $4.40 | |
| 14 | Kimi K2.5 Moonshot AI | 50.8 | moonshotai/Kimi-K2.5 | 38.6 | $2.85 | |
| 15 | DeepSeek V4 Flash 0731 DeepSeek | 50.7 | deepseek-ai/DeepSeek-V4-Flash-0731 | 51.8 | $0.28 | |
| 16 | GLM 4.7 Z.ai | 46.0 | zai-org/GLM-4.7 | 55.1 | $2.20 | |
| 17 | T Inkling Small Thinking Machines | 40.2 | thinkingmachines/Inkling-Small | 27.3 | $1.20 | |
| 18 | Gemma 4 31B Google | 26.0 | google/gemma-4-31B-it | 39.0 | $0.40 |
How this list is built
Generated from the dataset on every refresh. Nobody picks the order by hand.
Models on the Knowledge Index board that OpenRouter links to a public Hugging Face repository, by their Knowledge Index result. We do not check the license; read it on the repository page.
- OpenRouter, fetched
- LiveBench, fetched , board updated October 14, 2026
- MMLU Pro (Vals AI), fetched , board updated September 1, 2026
- Epoch AI benchmarking hub, fetched