KAware - Hybrid Composition: leaderboard

The knowing-acting score of KAware (2026) on tasks that combine tool calls with knowledge the model already holds: the harmonic mean of how well the agent states the needed tools and how well its actual tool calls match them, in percent; higher is better. KAware (2026) keeps tools available on every task, so an agent must judge whether it needs them.

Metric: The knowing-acting score (%, KAS): harmonic mean of knowing accuracy (Jaccard overlap between the tools the agent says the task needs and the gold tool set) and acting accuracy (Jaccard overlap between the tools it actually calls and the gold set), on the hybrid composition setting, on the KAware tool-use tasks, where tools are always available but the task may need them (external function, 310 tasks), mix them with parametric knowledge (hybrid composition, 368 tasks) or be solvable from knowledge alone (internal function, 398 tasks); higher is better. Source: arxiv.org. Saturation forecast: Around June 2027. 18 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro (Preview)64.64#113
2GPT-5.563.14#49
3Claude Sonnet 4.5 (Thinking)58.61#250 (Claude Sonnet 4.5)
4Qwen 3.5 397B A17B53.42#246
5Qwen 3 Max (Preview)53.22#447
6DeepSeek V4 Flash (Non-reasoning)51.6#258 (DeepSeek V4 Flash)
7GPT-551.33#167
8Qwen 3.5 397B A17B (Non-reasoning)51.12#246 (Qwen 3.5 397B A17B)
9O4 Mini (2025-04-16)50.32#365
10Qwen 3 235B A22B 2507 (Thinking)49.54#463
11DeepSeek V3.2 (Non-reasoning)49.26#406 (DeepSeek V3.2)
12Qwen 3 235B A22B 2507 Instruct48.23#619
13Claude Sonnet 4.547.62#250
14DeepSeek R1 052846.9#428
15Gemini 3 Flash (Preview)45.31#144

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=kaware-hybrid-composition · How It Works · Data refreshed daily, snapshot 2026-11-13.