SShogo Bench

TheAgentCompany

A simulated software company: browse, code, message coworkers.

Agent Index weight
0%
shown for context only
Models tested
14
from the benchmark’s leaderboard and reports
Normalised for the index
0 → 0, 1 → 100
Official site

Results

#ModelScoreNormalisedHarnessCost/taskDate
1DeepSeek V3.2 Supported
DeepSeek
42.9%source$0.40Oct 13, 2025
2Gemini 2.5 Flash Estimated
Google
41.1%source—Oct 13, 2025
3Claude Sonnet 4 Supported
Anthropic
33.1%source$1.63Jun 14, 2025
4Claude 3.7 Sonnet Supported
Anthropic
30.9%source$3.70Jun 14, 2025
5Gemini 2.5 Pro Supported
Google
30.3%source$4.23May 10, 2025
6Claude-Sonnet-3.5 Estimated
Anthropic
24.0%source$6.34Dec 17, 2024
7Gemini-2.0-Flash
Google
11.4%source$0.79Dec 17, 2024
8GPT-4o (2024-08-06) Supported
OpenAI
8.6%source$1.29Dec 17, 2024
9Llama 3.1-405B
Meta
7.4%source$3.21Dec 17, 2024
10Llama 3.1-70B
Meta
6.9%source$0.93Dec 17, 2024
11Qwen2.5-72B Estimated
Alibaba (Qwen)
5.7%source$1.53Dec 17, 2024
12Gemini 1.5 Pro (Feb 2024)
Google
3.4%source$6.78Dec 17, 2024
13Amazon Nova Pro
Amazon
1.7%source$1.55Dec 17, 2024
14Qwen2-72B
Alibaba (Qwen)
1.1%source$0.28Dec 17, 2024