// BENCHMARK

Frames benchmark

AI model leaderboard for the Frames benchmark. Compare how large language models score on Frames, see the full ranking, and understand what this AI benchmark measures. Tongyi DeepResearch currently leads with 90.6. Long-context reasoning benchmark — multi-hop questions over Wikipedia evidence requiring information aggregation.

Leaderboard

# Model Organization Score Variant Source
#1 Tongyi DeepResearch Alibaba-NLP 90.6 react-avg@3 official ↗
#2 Kimi K2 Thinking Moonshot 87 thinking-w-tools official ↗
#3 o3 OpenAI 84 react-cited-tongyi official ↗
#4 DeepSeek-V3.1 DeepSeek 83.7 react-cited-tongyi official ↗
#5 Claude Sonnet 4 Anthropic 80.7 react-cited-tongyi official ↗
#6 GLM 4.5 Zhipu AI 78.9 react-cited-tongyi official ↗
#7 Kimi K2 Moonshot 72 react-cited-tongyi official ↗
#8 Kimi Linear Moonshot 58.8 long-context-128k official ↗
#9 Kimi K2 Moonshot 58.1 k2-0905-w-tools-cited-k2-thinking official ↗

Frequently asked questions about Frames

What is the Frames benchmark?

Long-context reasoning benchmark — multi-hop questions over Wikipedia evidence requiring information aggregation.

How is the Frames benchmark scored?

Frames is scored using the accuracy metric, where a higher score is better. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on Frames?

As of the latest reported scores on GenAIList, Tongyi DeepResearch achieves the highest result on Frames with a score of 90.6.

Is a higher Frames score better?

Yes. On Frames a higher score indicates better performance, so models near the top of the leaderboard are the strongest.