Best LLMs for Science — MoonshotAI
The best MoonshotAI model for science is Kimi K3, scoring 93.5% on GPQA Diamond — graduate-level physics, chemistry and biology. Kimi K2.6 (91.1%) and Kimi K2.7 Code (89.6%) round out the top three.
- 1
kimi-k3ReasoningToolsJSON+157.1 intel · $3.00/M · 60 t/s93.5%GPQA - 2kimi-k2.6ReasoningToolsJSON+144.2 intel · $0.589/M · 92 t/s91.1%GPQA
- 3kimi-k2.7-codeReasoningToolsJSON+141.9 intel · $0.730/M · 185 t/s89.6%GPQA
- 4kimi-k2.5ReasoningToolsJSON+135.4 intel · $0.570/M · 49 t/s87.9%GPQA
- 5kimi-k2-thinkingReasoningToolsJSON32.7 intel · $0.600/M · 80 t/s83.8%GPQA
- 6kimi-k2-0905ToolsJSON23.5 intel · $0.600/M · 17 t/s76.7%GPQA
- 7kimi-k2Tools19.4 intel · $0.570/M · 27 t/s76.6%GPQA
How this is ranked
AI models ranked by GPQA Diamond — graduate-level physics, chemistry and biology questions written to be un-Googleable, so the score reflects reasoning from knowledge rather than retrieval. The best large language models for scientific research and hard-science work.
Frequently asked
What is the best MoonshotAI model for science?
The best MoonshotAI model for science is Kimi K3, scoring 93.5% on GPQA Diamond — graduate-level physics, chemistry and biology. Kimi K2.6 (91.1%) and Kimi K2.7 Code (89.6%) round out the top three.
What's a good alternative to Kimi K3?
Kimi K2.6 (91.1%) is the closest alternative on this metric, followed by Kimi K2.7 Code (89.6%). See the full ranking above for the tradeoffs.
How many MoonshotAI models are there?
modelgrep tracks 7 MoonshotAI models with live benchmarks, speed, latency and per-provider pricing, led on intelligence by Kimi K3. 7 of them qualify for this ranking.