benchgap
Agentic · tools

CyberGym leaderboard

As of 2026-11-21, the highest measured score on CyberGym is 95.1% by MiMo-V2.6-Flash. 180 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Muse Wonder Full97.7%estimated ± 6.0 pp, low confidence
2Kimi K3.1 Pro96.3%estimated ± 5.9 pp, low confidence
3MiMo-V2.6-Flash95.1%measured
4Claude Opus 594.8%estimated ± 4.5 pp, medium confidence
5Muse Wonder94.4%estimated ± 6.0 pp, low confidence
6MiMo-V2.6-Pro94.0%measured
7Claude Sonnet 5.592.7%estimated ± 6.0 pp, low confidence
8ERNIE 5.5 Flash Preview89.3%estimated ± 4.6 pp, medium confidence
9Hy488.3%estimated ± 4.6 pp, medium confidence
10Claude Haiku 5.588.2%estimated ± 6.0 pp, medium confidence
11GPT-5.5 Pro88.2%estimated ± 7.3 pp, medium confidence
12DeepSeek V4.1 Flash 101788.1%measured
13Gemini 4 Carbon87.9%estimated ± 4.6 pp, medium confidence
14Ling 3.1 Flash87.9%measured
15Hy4 flash87.6%estimated ± 4.6 pp, medium confidence
16LongCat-2.587.5%estimated ± 4.6 pp, medium confidence
17Claude Fable 5.287.4%estimated ± 4.6 pp, medium confidence
18Gemini 4 Argon87.4%estimated ± 4.6 pp, medium confidence
19Ultramarine 1 Flash87.1%estimated ± 4.6 pp, medium confidence
20Fugu Cyber86.9%measured
21GPT-5.4 Pro86.7%estimated ± 7.3 pp, medium confidence
22GPT-6 Cyber86.6%estimated ± 10.0 pp, low confidence
23Atria Dawn Preview86.5%measured
24Claude Mythos 5.286.5%estimated ± 10.0 pp, low confidence
25GPT-5.6 Cyber86.3%estimated ± 10.0 pp, medium confidence
26Gemini 3.8 Flash Cyber86.2%measured
27Claude Mythos 5.186.0%estimated ± 10.0 pp, medium confidence
28Gemini 3.8 Flash85.6%estimated ± 4.5 pp, high confidence
29Grok 4.685.4%estimated ± 3.1 pp, low confidence
30Muse Spark 1.385.4%estimated ± 3.1 pp, low confidence
31GPT-6 Astra84.9%estimated ± 4.6 pp, medium confidence
32LFM3-1.7B84.8%estimated ± 4.6 pp, medium confidence
33Step 5 Preview84.7%measured
34GLM-5.384.5%measured
35GPT-5.6 Sol84.5%measured
36Qwen3.8-27B84.4%estimated ± 3.1 pp, low confidence
37Claude Opus 5.584.4%estimated ± 4.6 pp, medium confidence
38Holo3-35B-A3B84.3%estimated ± 11.3 pp, low confidence
39Claude Fable 5.184.1%estimated ± 3.1 pp, low confidence
40GLM-5.3-Flash84.1%estimated ± 3.1 pp, low confidence
41Claude Mythos 583.8%measured
42Qwen3.8 Max Preview83.6%estimated ± 5.3 pp, medium confidence
43Kimi K383.6%estimated ± 3.1 pp, low confidence
44Muse Spark Open83.5%estimated ± 6.0 pp, low confidence
45Qwen3.8-Flash-Next83.4%estimated ± 3.1 pp, low confidence
46DeepSeek V4 Pro83.3%measured
47Gemini 3.5 Flash Cyber83.2%measured
48Ripple-65B Preview83.1%estimated ± 4.6 pp, medium confidence
49Claude Mythos Preview83.1%measured
50GPT-6.1 Sol83.0%estimated ± 4.6 pp, medium confidence
51MiniMax M3.1 Flash82.2%measured
52Grok 4.581.9%estimated ± 3.1 pp, low confidence
53GPT-6 Sol81.8%estimated ± 4.6 pp, medium confidence
54GPT-5.581.8%measured
55GPT-5.6 Terra81.8%measured
56Inkling 2 Small81.1%estimated ± 3.1 pp, low confidence
57Ling 3.0 Flash Fin80.2%estimated ± 3.1 pp, low confidence
58Ornith-1.5-397B80.0%estimated ± 6.0 pp, low confidence
59Claude Fable 579.9%estimated ± 3.1 pp, low confidence
60Gemini 3.7 Flash79.5%estimated ± 4.5 pp, high confidence
61Claude Sonnet 579.5%estimated ± 3.1 pp, low confidence
62GPT-6 Luna79.4%estimated ± 6.0 pp, medium confidence
63Muse Spark 1.279.3%estimated ± 5.3 pp, medium confidence
64GPT-5.479.0%measured
65LFM2.5-24B-A2B79.0%estimated ± 4.6 pp, medium confidence
66Mistral Large 478.9%estimated ± 6.0 pp, medium confidence
67Holo3-122B-A10B78.8%estimated ± 11.3 pp, low confidence
68Qwen3.8 Max78.7%estimated ± 4.6 pp, medium confidence
69Hy4 preview78.4%measured
70Claude Opus 4.878.3%estimated ± 5.3 pp, medium confidence
71GPT-5.6 Luna77.9%measured
72Grok 4.776.9%estimated ± 5.8 pp, low confidence
73dots3-note Preview76.7%estimated ± 7.3 pp, medium confidence
74DeepSeek V4 Flash 073176.7%measured
75GLM-5.275.8%estimated ± 5.3 pp, medium confidence
76UI-Mate-27B75.6%estimated ± 11.3 pp, low confidence
77Sage-A1.574.3%estimated ± 7.3 pp, medium confidence
78Solar Open 273.6%estimated ± 9.8 pp, medium confidence
79Claude Opus 4.773.1%measured
80Ling 3.0 Flash73.1%estimated ± 3.1 pp, low confidence
81Ling 3.0 Flash Sante72.0%estimated ± 7.3 pp, medium confidence
82Qwen3.5-27B71.3%estimated ± 5.8 pp, low confidence
83Quasar 438B70.7%estimated ± 5.3 pp, medium confidence
84Inkling 2 Mini70.2%estimated ± 7.3 pp, medium confidence
85MiniMax M369.2%estimated ± 5.3 pp, medium confidence
86Gemini 3.6 Flash68.7%estimated ± 5.3 pp, medium confidence
87GLM-5.168.7%measured
88Beam 1.0 501B-A23B FP866.9%estimated ± 7.3 pp, medium confidence
89Claude Opus 4.666.6%measured
90Gemini 3.5 Flash66.5%estimated ± 5.3 pp, medium confidence
91Gemini 3 Pro66.1%estimated ± 9.3 pp, medium confidence
92Apodex 1.165.6%estimated ± 6.0 pp, medium confidence
93Apodex 1.1 Mini65.6%estimated ± 6.0 pp, medium confidence
94Beam 1.0 501B-A23B NVFP465.3%estimated ± 7.3 pp, medium confidence
95Hy365.3%estimated ± 5.3 pp, medium confidence
96Hy3 Preview65.3%estimated ± 5.3 pp, medium confidence
97Agents-A165.2%estimated ± 7.3 pp, medium confidence
98Claude Sonnet 4.665.2%measured
99Beam 1.0 501B-A23B64.9%estimated ± 6.0 pp, medium confidence
100Inkling-Small64.7%estimated ± 5.3 pp, low confidence
101Inkling64.2%estimated ± 5.3 pp, low confidence
102Qwen3.7 Max63.9%estimated ± 5.3 pp, low confidence
103Nemotron 3.5 Super VL 120B A12B63.9%estimated ± 6.0 pp, medium confidence
104Nemotron 3.5 Super VL 120B A12B NVFP463.9%estimated ± 6.0 pp, medium confidence
105Ling 3.0 Flash VL63.8%estimated ± 6.0 pp, medium confidence
106MiMo-V2.5-Pro63.0%estimated ± 5.3 pp, low confidence
107Kimi K2.7 Code62.8%estimated ± 5.3 pp, low confidence
108Kimi K2.662.6%estimated ± 5.3 pp, low confidence
109Nemotron 3 Ultra62.2%estimated ± 5.3 pp, low confidence
110Qwen3.6-27B61.0%estimated ± 5.3 pp, low confidence
111Qwen3.7 Plus60.7%estimated ± 5.3 pp, low confidence
112GPT-5.4 mini60.6%estimated ± 5.3 pp, low confidence
113Solar Pro 460.2%estimated ± 6.0 pp, medium confidence
114GPT-5.4 nano59.2%estimated ± 5.3 pp, low confidence
115Muse Spark 1.159.0%measured
116Grok 4.358.8%estimated ± 5.3 pp, low confidence
117MiniMax M2.758.5%estimated ± 5.3 pp, low confidence
118Nemotron 3.5 Bolt 74B-A9B BF1657.9%estimated ± 7.3 pp, medium confidence
119Gemini 3.5 Flash-Lite57.8%estimated ± 5.3 pp, low confidence
120Qwen3.6-35B-A3B57.1%estimated ± 5.3 pp, low confidence
121LLaDA2.2-flash56.7%estimated ± 9.8 pp, medium confidence
122GPT-5.3 Codex56.7%estimated ± 9.3 pp, medium confidence
123Qwen3.5 397B56.6%estimated ± 7.3 pp, medium confidence
124Gemini 3 Flash55.3%estimated ± 9.3 pp, medium confidence
125Ornith-1.5-35B-A3B54.9%estimated ± 7.3 pp, medium confidence
126Agents-A1-4B53.9%estimated ± 7.3 pp, medium confidence
127Step 3.7 Flash53.7%estimated ± 6.0 pp, medium confidence
128Muse Glimmer 30B53.6%estimated ± 5.3 pp, low confidence
129GLM-4.753.5%estimated ± 6.0 pp, medium confidence
130Gemini 3.1 Pro53.5%estimated ± 5.3 pp, low confidence
131Qwen3.5-122B-A10B53.0%estimated ± 5.3 pp, low confidence
132Mistral Medium 3.5 128B52.7%estimated ± 5.3 pp, low confidence
133GPT-5.252.7%estimated ± 7.3 pp, medium confidence
134UI-Mate-9B52.4%estimated ± 11.3 pp, low confidence
135Qwen3.6 Plus51.7%estimated ± 6.0 pp, low confidence
136Nemotron 3.5 Bolt 74B-A9B50.9%estimated ± 6.0 pp, low confidence
137Gemma 4 31B50.7%estimated ± 5.3 pp, low confidence
138Claude Opus 4.550.6%measured
139GPT-OSS 120B50.3%estimated ± 5.3 pp, low confidence
140Nemotron 3.5 Lightning 30B A3B NVFP450.3%estimated ± 5.3 pp, low confidence
141A.X K249.3%estimated ± 6.0 pp, low confidence
142Granite 4.2 8B48.4%estimated ± 5.3 pp, low confidence
143Command A+48.3%estimated ± 5.3 pp, low confidence
144Gemini 2.5 Pro48.3%estimated ± 5.3 pp, low confidence
145GPT-5.2-Codex48.0%estimated ± 9.3 pp, medium confidence
146Claude 4.1 Opus47.6%estimated ± 11.4 pp, low confidence
147Mistral Large 347.4%estimated ± 5.3 pp, low confidence
148Qwen3.5-35B-A3B47.2%estimated ± 7.3 pp, low confidence
149Mistral Small 446.7%estimated ± 5.3 pp, low confidence
150GPT-OSS 20B46.6%estimated ± 5.3 pp, low confidence
151Trinity-Large-Preview46.4%estimated ± 5.3 pp, low confidence
152Trinity-Large-Thinking46.4%estimated ± 5.3 pp, low confidence
153Nemotron 3 Nano 30B46.3%estimated ± 5.3 pp, low confidence
154DeepSeek V346.2%estimated ± 5.3 pp, low confidence
155Celeris-146.1%estimated ± 5.3 pp, low confidence
156Llama 4 Maverick46.0%estimated ± 5.3 pp, low confidence
157Llama 4 Scout46.0%estimated ± 5.3 pp, low confidence
158GPT-545.9%estimated ± 6.0 pp, low confidence
159Gemma 3 27B45.7%estimated ± 5.3 pp, low confidence
160GPT-5.1-Codex44.9%estimated ± 9.3 pp, low confidence
161Grok Build 0.144.2%estimated ± 9.3 pp, low confidence
162Muse Spark43.5%measured
163Claude Sonnet 4.543.2%estimated ± 9.3 pp, low confidence
164GLM-543.2%measured
165Ornith-1.5-9B42.2%estimated ± 7.3 pp, low confidence
166Qwen3.5 Plus42.2%estimated ± 11.4 pp, low confidence
167Grok 4.1 Fast41.6%estimated ± 9.3 pp, low confidence
168MiMo-V2.541.0%estimated ± 9.3 pp, low confidence
169Kimi K2.539.0%estimated ± 6.0 pp, low confidence
170Spark-X2.6-4B38.9%estimated ± 7.3 pp, low confidence
171Claude Haiku 4.537.9%estimated ± 11.4 pp, low confidence
172GPT-5.137.7%estimated ± 6.0 pp, low confidence
173Qwen3 Max36.7%estimated ± 9.3 pp, low confidence
174K-EXAONE 2.036.0%estimated ± 6.0 pp, low confidence
175Sarvam 105B35.4%estimated ± 7.3 pp, low confidence
176Grok 434.8%estimated ± 9.3 pp, low confidence
177LongCat-Flash-Lite-Sparse34.5%estimated ± 7.3 pp, low confidence
178Claude 4 Sonnet31.4%estimated ± 9.3 pp, low confidence
179Gemini 3.1 Flash-Lite29.9%estimated ± 9.3 pp, low confidence
180Grok 4.2029.7%estimated ± 9.3 pp, low confidence
181MiMo-V2-Pro27.7%estimated ± 9.3 pp, low confidence
182MiniCPM5-2B25.9%estimated ± 6.0 pp, low confidence
183Sarvam 30B23.2%estimated ± 7.3 pp, low confidence
184GLM-5V-Turbo20.6%estimated ± 9.3 pp, low confidence
185DeepSeek V3.219.3%estimated ± 9.3 pp, low confidence
186MiMo-V2-Flash16.8%estimated ± 6.0 pp, low confidence
187Granite 4.2 30B10.0%estimated ± 6.0 pp, low confidence
188Gemma 4 26B A4B8.8%estimated ± 6.0 pp, low confidence
189Ling 3.0 Tiny8.2%estimated ± 6.0 pp, low confidence
190Claude 3.5 Haiku0.0%estimated ± 6.0 pp, low confidence
191DeepSeek V3 03240.0%estimated ± 6.0 pp, low confidence
192Gemma 4 12B0.0%estimated ± 6.0 pp, low confidence
193Gemma 4 E2B0.0%estimated ± 6.0 pp, low confidence
194Gemma 4 E4B0.0%estimated ± 6.0 pp, low confidence
195GPT-4.1 mini0.0%estimated ± 6.0 pp, low confidence
196GPT-4.1 nano0.0%estimated ± 6.0 pp, low confidence
197GPT-4o0.0%estimated ± 6.0 pp, low confidence
198GPT-4o mini0.0%estimated ± 6.0 pp, low confidence
199Granite 4.2 3B0.0%estimated ± 6.0 pp, low confidence
200K-Exaone0.0%estimated ± 6.0 pp, low confidence
201LFM2.5-2.6B0.0%estimated ± 6.0 pp, low confidence
202Ling 2.6 Flash0.0%estimated ± 6.0 pp, low confidence
203Llama 3.3 70B0.0%estimated ± 6.0 pp, low confidence
204Mercury 2.50.0%estimated ± 6.0 pp, low confidence
205Mistral Small 3.10.0%estimated ± 6.0 pp, low confidence
206Nemotron 3 Nano Omni 30B A3B0.0%estimated ± 6.0 pp, low confidence
207North Mini Code0.0%estimated ± 6.0 pp, low confidence
208Solar Pro 30.0%estimated ± 6.0 pp, low confidence
209Ultravox v0.6 Llama 3.3 70B0.0%estimated ± 6.0 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General