KAware - Internal Function: leaderboard
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 internal function 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 December 2026. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 72.22 |
| 2 | O4 Mini (2025-04-16) | 57.41 |
| 3 | GPT-5.5 | 44.15 |
| 4 | Gemini 3 Pro (Preview) | 38.83 |
| 5 | GPT-4.1 | 35.98 |
| 6 | DeepSeek R1 0528 | 35.68 |
| 7 | Gemini 3 Flash (Preview) | 33.24 |
| 8 | GPT-4o (2024-11-20) | 31.98 |
| 9 | Qwen 3.5 397B A17B | 28.67 |
| 10 | Qwen 3.5 397B A17B (Non-reasoning) | 28.13 |
| 11 | Claude Sonnet 4.5 | 25.7 |
| 12 | DeepSeek V4 Flash (Non-reasoning) | 21.89 |
| 13 | Claude Sonnet 4.5 (Thinking) | 21.84 |
| 14 | Qwen 3 235B A22B 2507 (Thinking) | 21.4 |
| 15 | DeepSeek V3.2 (Non-reasoning) | 16.54 |
Interactive version: theaggregate.ai/benchmark?slug=kaware-internal-function · How It Works · Data refreshed daily, snapshot 2026-09-29.