Can Agents Price a Reaction? Evaluating LLMs o: leaderboard
Large Language Models (LLMs) have become increasingly capable as tool-using agents, with benchmarks spanning diverse general agentic tasks. Yet rigorous evaluation of scientific tool use remains limited. In chemistry, recent agents can plan syntheses and invoke domain-specific tools, but evaluations often rely on curated demonstrations, expert assessment, or LLM-as-judge scoring rather than exact,
Metric: C.10 Clean (self-reported). Source: benchmarklist.com. Saturation forecast: Around July 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 37 | #102 |
| 2 | Qwen 3.5 Plus | 34.6 | #125 |
| 3 | GPT-5 | 24.7 | #98 |
| 4 | Kimi K2.5 | 24.7 | #144 |
| 5 | Claude Sonnet 4.6 | 24.3 | #90 |
| 6 | Gemini 2.5 Flash | 21.8 | #245 |
| 7 | GPT-5.4 Mini | 5.8 | #212 |
| 8 | Qwen3-14B-Base | 4.9 | #623 |
| 9 | Qwen 3 235B A22B | 3.7 | #313 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=can-agents-price-a-reaction-evaluating-llms-o · How It Works · Data refreshed daily, snapshot 2026-10-20.