# Scale Labs SEAL Agentic Tool Use (Chat) (SEAL Agentic Tool Use (Chat))

> ToolComp-Chat: 485-prompt benchmark subset that needs models to chain a Python interpreter and web search calls to reach a final answer. BenchLM shows it as display-only reference data, excluded from rankings. Scale has not added a row to this board since March 2025.

Canonical page: https://benchlm.ai/benchmarks/scale-tool-use-chat

- Category: [Agentic](/agentic)
- Last updated: November 12, 2026 snapshot

## About SEAL Agentic Tool Use (Chat)

- Year: 2026
- Tasks: ToolComp prompts, Chat subset
- Format: Accuracy (%)
- Difficulty: External agent and model evaluation
- Paper: [Scale Labs leaderboard](https://labs.scale.com/leaderboard/tool_use_chat)

ToolComp prompts require several dependent tool calls, with golden answer chains and step-level labels. The Chat subset uses two common chatbot tools. The score is final-answer accuracy in percent. BenchLM mirrors 35 published rows captured on November 12, 2026 snapshot.

SEAL Agentic Tool Use (Chat) is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (35 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [o3-mini (high)](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 63.45% |
| 2 | [Gemini 2.5 Pro Experimental (March 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | Google | 62.43% |
| 3 | [o3-mini (medium)](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 62.42% |
| 4 | [o1 Pro (March 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 61.42% |
| 5 | [DeepSeek-R1](/models/deepseek-r1) | DeepSeek | 60.91% |
| 6 | [o1 (December 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 60.41% |
| 7 | [DeepSeek-V3 (December 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | Deepseek | 58.55% |
| 8 | [Gemini 2.0 Pro Experimental (February 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | Google | 57.86% |
| 9 | [Gemini 2.0 Flash Thinking Experimental (January 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | Google | 57.36% |
| 10 | [GPT-4o (August 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 56.85% |
| 11 | [GPT-4.5 Preview (February 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 56.34% |
| 12 | [Claude 3.7 Sonnet (February 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | Anthropic | 56.25% |
| 13 | [Claude 3.5 Sonnet (June 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | Anthropic | 56.06% |
| 14 | [Claude 3.7 Sonnet Thinking (February 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | Anthropic | 55.32% |
| 15 | [o1-preview](/models/o1-preview) | OpenAI | 55.10% |
| 16 | [DeepSeek V3 (March 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | Deepseek | 54.31% |
| 17 | [Gemini 2.0 Flash Experimental (December 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | Google | 53.29% |
| 18 | [GPT-4 Turbo Preview](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 53.03% |
| 19 | [Gemini 1.5 Pro (August 27, 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | Google | 51.27% |
| 20 | [Gemini 2.0 Flash Lite Preview (February 2025)](https://labs.scale.com/leaderboard/tool_use_chat) | Google | 50.25% |
| 21 | [GPT-4o (May 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 49.50% |
| 22 | [Nova Pro](/models/nova-pro) | Amazon | 49.23% |
| 23 | [Claude 3 Opus](/models/claude-3-opus) | Anthropic | 48.49% |
| 24 | [Gemini 2.0 Flash](/models/gemini-2-0-flash) | Google | 48.22% |
| 25 | [Nova LIte](https://labs.scale.com/leaderboard/tool_use_chat) | Amazon | 41.66% |
| 26 | [Nova Mirco](https://labs.scale.com/leaderboard/tool_use_chat) | Amazon | 40.97% |
| 27 | [Claude 3 Sonnet](https://labs.scale.com/leaderboard/tool_use_chat) | Anthropic | 40.40% |
| 28 | [Mistral Large 2](/models/mistral-large-2) | Mistral | 40.40% |
| 29 | [Llama 3.1 405B Instruct](https://labs.scale.com/leaderboard/tool_use_chat) | Meta | 40.10% |
| 30 | [GPT-4](https://labs.scale.com/leaderboard/tool_use_chat) | OpenAI | 37.88% |
| 31 | [Gemini 1.5 Pro (May 2024)](https://labs.scale.com/leaderboard/tool_use_chat) | Google | 35.50% |
| 32 | [Llama 3.1 70B Instruct](https://labs.scale.com/leaderboard/tool_use_chat) | Meta | 33.50% |
| 33 | [GPT-4o mini](/models/gpt-4o-mini) | OpenAI | 32.83% |
| 34 | [Command R+](https://labs.scale.com/leaderboard/tool_use_chat) | Cohere | 20.20% |
| 35 | [Llama 3.1 8B Instruct](https://labs.scale.com/leaderboard/tool_use_chat) | Meta | 6.09% |

## FAQ

### What does SEAL Agentic Tool Use (Chat) measure?

ToolComp-Chat: 485-prompt benchmark subset that needs models to chain a Python interpreter and web search calls to reach a final answer. BenchLM shows it as display-only reference data, excluded from rankings. Scale has not added a row to this board since March 2025.

### Which model leads the published SEAL Agentic Tool Use (Chat) snapshot?

o3-mini (high) currently leads the published SEAL Agentic Tool Use (Chat) snapshot with a score of 63.45%.

### How many models are evaluated on SEAL Agentic Tool Use (Chat)?

The November 12, 2026 snapshot contains 35 AI models.

### Does SEAL Agentic Tool Use (Chat) affect BenchLM's overall score?

Not directly. SEAL Agentic Tool Use (Chat) is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.