SShogo Bench

MirrorCode

Reimplement a program from its behaviour alone.

Agent Index weight
0%
shown for context only
Models tested
41
run by Epoch AI on a shared harness
Normalised for the index
0 → 0, 1 → 100
Official site

Results

#ModelScoreNormalisedHarnessCost/taskDate
1Claude Sonnet 5.5 Estimated
Anthropic
97.8%source—Oct 5, 2026
2Claude Opus 5.5 Supported
Anthropic
95.6%source—Oct 30, 2026
3Claude Fable 5.1 Supported
Anthropic
93.3%source—Sep 9, 2026
4Claude Opus 5 Supported
Anthropic
87.8%source—Oct 20, 2026
5Claude Haiku 5.5 Estimated
Anthropic
78.9%source—Oct 21, 2026
6GPT-6.1 Terra Estimated
OpenAI
76.7%source—Oct 30, 2026
7Claude Fable 5 Supported
Anthropic
70.0%source—Oct 30, 2026
8GPT-6.2 Astra
OpenAI
67.8%source—Oct 27, 2026
9Claude Opus 4.8 Supported
Anthropic
65.6%source—Oct 21, 2026
10DeepSeek V4.1 Flash Estimated
DeepSeek
64.4%source—Oct 21, 2026
11Claude Sonnet 5 Supported
Anthropic
63.3%source—Oct 21, 2026
12GPT-6.1 Sol Estimated
OpenAI
61.1%source—Oct 5, 2026
13GPT-6 Astra Supported
OpenAI
60.0%source—Oct 29, 2026
14Gemini 4 Argon Supported
Google
58.9%source—Oct 14, 2026
15Grok 4.7 Estimated
xAI
54.4%source—Oct 29, 2026
16Kimi K3 Supported
Moonshot AI
54.4%source—Sep 3, 2026
17Muse Spark 1.3 Estimated
Meta
54.4%source—Oct 20, 2026
18Qwen3.8 Max Preview Estimated
Alibaba (Qwen)
52.2%source—Oct 21, 2026
19GPT-6 Sol Estimated
OpenAI
48.9%source—Oct 29, 2026
20Qwen3.8 Max 0902
Alibaba (Qwen)
47.8%source—Oct 21, 2026
21Muse Wonder Estimated
Meta
45.6%source—Oct 14, 2026
22Claude Opus 4.7 Supported
Anthropic
43.3%source—Oct 30, 2026
23Kimi K3.1 Pro
Moonshot AI
43.3%source—Nov 17, 2026
24Gemini 3.7 Flash Estimated
Google
41.1%source—Oct 21, 2026
25GPT-5.6 Terra Estimated
OpenAI
41.1%source—Oct 20, 2026
26Grok 4.6 Supported
xAI
41.1%source—Oct 21, 2026
27Mistral Large 4
Mistral
40.0%source—Nov 2, 2026
28Gemini 3.8 Flash Supported
Google
38.9%source—Sep 2, 2026
29GLM-5.2 Supported
Z.ai (Zhipu)
36.7%source—Oct 21, 2026
30Claude Opus 4.6 Supported
Anthropic
33.3%source—Oct 21, 2026
31DeepSeek V4 Pro 0813
DeepSeek
33.3%source—Oct 20, 2026
32GPT-5.2 Supported
OpenAI
32.2%source—Oct 21, 2026
33Grok 4.5 Estimated
xAI
32.2%source—Oct 21, 2026
34GPT-6 Luna
OpenAI
26.7%source—Oct 21, 2026
35Gemini 3.6 Flash Estimated
Google
23.3%source—Oct 21, 2026
36GPT-5.4 Supported
OpenAI
23.3%source—Oct 29, 2026
37GPT-5.5 Supported
OpenAI
23.3%source—Oct 29, 2026
38GPT-5.6 Luna Estimated
OpenAI
21.1%source—Oct 21, 2026
39GPT-5.6 Sol Supported
OpenAI
20.0%source—Aug 10, 2026
40Gemini 3.1 Pro Preview Supported
Google
8.9%source—Aug 10, 2026
41GPT-5 Supported
OpenAI
4.4%source—Oct 21, 2026