SShogo Bench

CL-bench

Learn new rules, procedures and domain knowledge from context, then apply them.

Agent Index weight
5%
in the Security & context learning pillar
Models tested
19
from the benchmark’s leaderboard and reports
Normalised for the index
0 → 0, 1 → 100
Official site

Results

#ModelScoreNormalisedHarnessCost/taskDate
1GPT-5.4 Supported
OpenAI
27.9%source——
2GPT-5.1 Supported
OpenAI
23.7%source——
3Grok 4.20 (Non-Reasoning) Estimated
xAI
22.2%source——
4Claude Opus 4.5 (latest) Supported
Anthropic
21.1%source——
5Gemini 3.1 Pro Preview Supported
Google
20.8%source——
6Claude Opus 4.6 Supported
Anthropic
20.7%source——
7Qwen3.6 Plus Estimated
Alibaba (Qwen)
20.3%source——
8Qwen3.5 Plus Estimated
Alibaba (Qwen)
19.8%source——
9Kimi K2.5 Supported
Moonshot AI
19.3%source——
10GLM-5 Supported
Z.ai (Zhipu)
18.7%source——
11GPT-5.2 Supported
OpenAI
18.2%source——
12o3 Supported
OpenAI
17.8%source——
13Kimi K2 Thinking Estimated
Moonshot AI
17.6%source——
14GLM-4.7 Estimated
Z.ai (Zhipu)
15.9%source——
15Gemini 3 Pro Preview Supported
Google
15.8%source——
16MiMo V2 Pro Free
Xiaomi
15.7%source——
17Qwen3 Max Estimated
Alibaba (Qwen)
14.5%source——
18DeepSeek V3.2 Supported
DeepSeek
13.2%source——
19MiniMax-M2.5 Estimated
MiniMax
11.4%source——

For models without a CL-bench result, the index estimates one from Epoch’s Capabilities Index using a logistic fit on 17 models (residual SD 0.17 on the logit scale). Estimates are labelled and widen the model’s range.