Agentic & Tool Use benchmark
TheAgentCompany leaderboard
As of October 2026, DeepSeek-V3.2-Exp has the highest published TheAgentCompany score on Noometry at 42.9%, out of 14 models with results.
Last verified
About TheAgentCompany
Workplace tasks in a simulated software company with browsers, code, chat and documents.
- Category
- Agentic & Tool Use
- Introduced
- 2024
- Size
- 175 tasks
- Format
- Agent in a sandbox
- Unit
- Percent (random guessing ≈ 0%)
- Official site
- the-agent-company.com
Top 14 models
- DeepSeek-V3.2-Exp 42.9%
- Gemini 2.5 Flash 41.1%
- Claude Sonnet 4 33.1%
- Claude 3.7 Sonnet 30.9%
- Gemini 2.5 Pro 30.3%
- Claude 3.5 Sonnet 24%
- Gemini 2.0 Flash (Feb 2025) 11.4%
- GPT-4o 8.6%
- Llama 3.1-405B 7.4%
- Llama 3.1-70B 6.9%
- Qwen2.5 72B Instruct 5.7%
- Gemini 1.5 Pro (May 2024) 3.4%
- Amazon Nova Pro 1.7%
- Qwen2-72B 1.1%
Sponsored placements are available on pages like this one. Advertise on Noometry
All results
| # | Model | Provider | Score | Setting | Source | Date |
|---|---|---|---|---|---|---|
| 1 | DeepSeek-V3.2-Exp |  DeepSeek | 42.9% | Epoch AI | ||
| 2 | Gemini 2.5 Flash |  Google | 41.1% | Epoch AI | ||
| 3 | Claude Sonnet 4 | Anthropic | 33.1% | Epoch AI | ||
| 4 | Claude 3.7 Sonnet | Anthropic | 30.9% | Epoch AI | ||
| 5 | Gemini 2.5 Pro | Google | 30.3% | Epoch AI | ||
| 6 | Claude 3.5 Sonnet | Anthropic | 24% | Epoch AI | ||
| 7 | Gemini 2.0 Flash (Feb 2025) | Google | 11.4% | Epoch AI | ||
| 8 | GPT-4o | OpenAI | 8.6% | Epoch AI | ||
| 9 | Llama 3.1-405B |  Meta | 7.4% | Epoch AI | ||
| 10 | Llama 3.1-70B | Meta | 6.9% | Epoch AI | ||
| 11 | Qwen2.5 72B Instruct |  Alibaba (Qwen) | 5.7% | Epoch AI | ||
| 12 | Gemini 1.5 Pro (May 2024) | Google | 3.4% | Epoch AI | ||
| 13 | Amazon Nova Pro | Amazon | 1.7% | Epoch AI | ||
| 14 | Qwen2-72B | Alibaba (Qwen) | 1.1% | Epoch AI |
Compare the leaders
Frequently asked questions
What does TheAgentCompany measure?
Workplace tasks in a simulated software company with browsers, code, chat and documents.
Which model has the highest TheAgentCompany score?
As of October 2026, DeepSeek-V3.2-Exp has the highest published TheAgentCompany score on Noometry at 42.9%, out of 14 models with results.
What is the best open-weight model on TheAgentCompany?
DeepSeek-V3.2-Exp has the highest TheAgentCompany accuracy among open-weight models at 42.9%, ranking 1 of 14 overall.