benchgap
Agentic · tools

MCP Atlas leaderboard

As of 2026-11-21, the highest measured score on MCP Atlas is 89.2% by Muse Spark Open. 163 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Qwen4 Max97.8%estimated ± 9.1 pp, low confidence
2Muse Wonder Full93.5%estimated ± 8.6 pp, low confidence
3Muse Wonder91.0%estimated ± 8.6 pp, medium confidence
4Muse Spark Open89.2%measured
5Muse Spark 1.188.1%measured
6Ultramarine 1 Flash86.7%estimated ± 8.6 pp, medium confidence
7Claude Opus 585.8%measured
8Step 5 Preview85.6%measured
9Qwen3.8 Max85.5%estimated ± 6.6 pp, low confidence
10GLM-5.385.1%estimated ± 6.0 pp, low confidence
11Inkling 2 Small85.0%measured
12Claude Fable 5.285.0%estimated ± 7.6 pp, low confidence
13Hy484.7%estimated ± 6.6 pp, low confidence
14Kimi K3.1 Pro84.7%estimated ± 6.6 pp, low confidence
15Hy4 flash84.7%estimated ± 6.6 pp, low confidence
16Claude Fable 584.7%estimated ± 5.8 pp, low confidence
17ERNIE 5.5 Flash Preview84.6%measured
18Apodex 1.184.5%estimated ± 6.6 pp, low confidence
19Claude Sonnet 5.584.2%estimated ± 7.5 pp, medium confidence
20Kimi K384.2%measured
21MiMo-V2.6-Pro83.8%estimated ± 7.5 pp, medium confidence
22Atria Dawn Preview83.8%estimated ± 6.6 pp, low confidence
23Hy4 preview83.7%measured
24Gemini 3.5 Flash83.6%measured
25Gemini 4 Argon83.5%estimated ± 10.4 pp, low confidence
26MiniMax M3.1 Flash83.4%estimated ± 10.4 pp, low confidence
27Qwen3.8-27B83.1%estimated ± 6.0 pp, low confidence
28LongCat-2.583.1%measured
29Ling 3.1 Flash83.0%measured
30Inkling 2 Mini82.7%measured
31dots3-note Preview82.6%estimated ± 6.6 pp, low confidence
32Claude Fable 5.182.5%estimated ± 6.0 pp, low confidence
33GLM-5.3-Flash82.5%estimated ± 6.0 pp, low confidence
34GPT-6 Sol82.3%estimated ± 6.8 pp, medium confidence
35Claude Haiku 5.582.2%estimated ± 10.7 pp, low confidence
36MiMo-V2.6-Flash82.2%estimated ± 7.5 pp, medium confidence
37Claude Opus 4.882.2%measured
38GPT-6.1 Sol82.1%estimated ± 10.7 pp, low confidence
39GPT-5.6 Sol81.6%estimated ± 5.8 pp, low confidence
40Beam 1.0 501B-A23B FP881.5%measured
41GPT-6 Astra81.5%estimated ± 6.8 pp, medium confidence
42GPT-6 Luna81.0%estimated ± 10.7 pp, low confidence
43Qwen3.8-Flash-Next80.9%estimated ± 6.0 pp, low confidence
44Gemini 3.8 Flash80.8%estimated ± 5.8 pp, low confidence
45Beam 1.0 501B-A23B80.8%measured
46Beam 1.0 501B-A23B NVFP480.5%measured
47Ornith-1.5-397B80.0%measured
48Apodex 1.1 Mini79.8%estimated ± 6.6 pp, low confidence
49DeepSeek V4.1 Flash 101779.8%estimated ± 6.8 pp, medium confidence
50Inkling-Small79.6%measured
51Fugu Cyber79.5%estimated ± 11.1 pp, low confidence
52GPT-5.5 Pro79.5%estimated ± 10.3 pp, low confidence
53GPT-5.4 Pro79.5%estimated ± 10.3 pp, low confidence
54Claude Mythos 579.4%estimated ± 10.3 pp, low confidence
55Gemini 3.8 Flash Cyber79.4%estimated ± 11.1 pp, low confidence
56Mistral Large 479.3%estimated ± 10.3 pp, low confidence
57Sage-A1.579.1%estimated ± 10.3 pp, low confidence
58Gemini 3.5 Flash Cyber79.0%estimated ± 11.1 pp, low confidence
59Claude Mythos Preview79.0%estimated ± 11.1 pp, low confidence
60Muse Spark 1.378.9%estimated ± 5.8 pp, low confidence
61Qwen3.8 Max Preview78.7%estimated ± 7.8 pp, low confidence
62Muse Spark 1.278.7%estimated ± 7.8 pp, low confidence
63Quasar 438B78.7%estimated ± 7.8 pp, low confidence
64Gemini 3.6 Flash78.7%estimated ± 7.8 pp, low confidence
65Hy378.3%estimated ± 7.8 pp, low confidence
66Hy3 Preview78.3%estimated ± 7.8 pp, low confidence
67Grok 4.577.9%estimated ± 6.0 pp, low confidence
68Nemotron 3.5 Super VL 120B A12B77.8%estimated ± 10.3 pp, low confidence
69Claude Sonnet 577.7%estimated ± 5.8 pp, low confidence
70Gemini 3.7 Flash77.7%estimated ± 5.8 pp, low confidence
71GPT-5.6 Terra77.7%estimated ± 5.8 pp, low confidence
72Nemotron 3.5 Super VL 120B A12B NVFP477.4%estimated ± 10.3 pp, low confidence
73Claude Opus 4.777.3%measured
74GLM-5.276.8%measured
75Grok 4.676.5%estimated ± 5.8 pp, low confidence
76Qwen3.7 Max76.4%measured
77Ornith-1.0-397B76.2%estimated ± 13.5 pp, low confidence
78Kimi K2.7 Code76.0%measured
79Ling 3.0 Flash Sante75.6%measured
80BTL-375.5%estimated ± 14.2 pp, low confidence
81Muse Glimmer 30B75.5%measured
82Agents-A175.3%estimated ± 10.3 pp, low confidence
83GPT-5.575.3%measured
84Ling 3.0 Flash Fin74.7%estimated ± 6.0 pp, low confidence
85Grok 4.774.6%estimated ± 6.8 pp, medium confidence
86GPT-5.6 Luna74.2%estimated ± 5.8 pp, low confidence
87MiniMax M374.2%measured
88Inkling74.1%measured
89BTL-473.7%estimated ± 14.2 pp, low confidence
90DeepSeek V4 Pro73.6%measured
91Claude Opus 4.673.4%estimated ± 5.8 pp, low confidence
92Qwen3.7 Plus73.2%measured
93Ling 3.0 Flash VL73.1%estimated ± 10.7 pp, low confidence
94Claude Opus 4.6 (Adaptive)72.4%estimated ± 9.1 pp, medium confidence
95Claude Sonnet 4.672.3%estimated ± 5.8 pp, low confidence
96GLM-5.171.8%measured
97Ripple-65B Preview71.7%measured
98Qwen 3.6 Max (preview)71.1%estimated ± 12.3 pp, low confidence
99GPT-5.470.6%measured
100Ornith-1.5-35B-A3B70.2%measured
101Claude Opus 5.570.2%estimated ± 6.8 pp, medium confidence
102MiMo-V2.5-Pro70.2%estimated ± 6.8 pp, medium confidence
103Gemma 4 31B69.1%estimated ± 6.8 pp, medium confidence
104DeepSeek V4 Flash 073169.0%measured
105Laguna S 2.167.7%estimated ± 7.5 pp, medium confidence
106Qwen3.5-27B66.9%estimated ± 6.8 pp, medium confidence
107Nemotron 3 Ultra65.7%estimated ± 7.8 pp, low confidence
108Ling 3.0 Flash65.5%measured
109Ornith-1.0-35B65.1%estimated ± 13.5 pp, low confidence
110Qwen3.6-27B64.7%estimated ± 7.8 pp, low confidence
111MiMo-V2.564.6%estimated ± 12.9 pp, low confidence
112Grok 4.364.6%estimated ± 7.8 pp, low confidence
113MiniMax M2.764.6%estimated ± 7.8 pp, low confidence
114Gemini 3.5 Flash-Lite64.6%estimated ± 7.8 pp, low confidence
115Gemini 3.1 Pro64.6%estimated ± 7.8 pp, low confidence
116Qwen3.5-122B-A10B64.6%estimated ± 7.8 pp, low confidence
117Mistral Medium 3.5 128B64.6%estimated ± 7.8 pp, low confidence
118GPT-OSS 120B64.6%estimated ± 7.8 pp, low confidence
119Nemotron 3.5 Lightning 30B A3B NVFP464.6%estimated ± 7.8 pp, low confidence
120Granite 4.2 8B64.6%estimated ± 7.8 pp, low confidence
121Command A+64.6%estimated ± 7.8 pp, low confidence
122Gemini 2.5 Pro64.6%estimated ± 7.8 pp, low confidence
123Mistral Large 364.6%estimated ± 7.8 pp, low confidence
124Mistral Small 464.6%estimated ± 7.8 pp, low confidence
125GPT-OSS 20B64.6%estimated ± 7.8 pp, low confidence
126Nemotron 3 Nano 30B64.6%estimated ± 7.8 pp, low confidence
127Trinity-Large-Preview64.6%estimated ± 7.8 pp, low confidence
128Trinity-Large-Thinking64.6%estimated ± 7.8 pp, low confidence
129Celeris-164.6%estimated ± 7.8 pp, low confidence
130DeepSeek V364.6%estimated ± 7.8 pp, low confidence
131Gemma 3 27B64.6%estimated ± 7.8 pp, low confidence
132Llama 4 Maverick64.6%estimated ± 7.8 pp, low confidence
133Llama 4 Scout64.6%estimated ± 7.8 pp, low confidence
134GPT-5.263.6%estimated ± 8.6 pp, medium confidence
135GPT-5.3 Codex63.0%estimated ± 8.6 pp, medium confidence
136Muse Spark62.8%estimated ± 10.7 pp, low confidence
137Qwen3.6-35B-A3B62.8%measured
138Spark-X2.6-4B62.4%measured
139Solar Pro 461.4%measured
140A.X K260.8%estimated ± 10.7 pp, low confidence
141Solar Open 258.2%measured
142GPT-5.4 mini57.7%measured
143GPT-5.157.2%estimated ± 10.7 pp, low confidence
144K-EXAONE 2.057.0%estimated ± 10.7 pp, low confidence
145MiniCPM5-2B56.6%estimated ± 10.7 pp, low confidence
146MiMo-V2-Flash56.5%estimated ± 10.7 pp, low confidence
147Granite 4.2 30B56.5%estimated ± 10.7 pp, low confidence
148Gemma 4 26B A4B56.5%estimated ± 10.7 pp, low confidence
149Ling 3.0 Tiny56.5%estimated ± 10.7 pp, low confidence
150Claude 3.5 Haiku56.5%estimated ± 10.7 pp, low confidence
151DeepSeek V3 032456.5%estimated ± 10.7 pp, low confidence
152Gemma 4 12B56.5%estimated ± 10.7 pp, low confidence
153Gemma 4 E2B56.5%estimated ± 10.7 pp, low confidence
154Gemma 4 E4B56.5%estimated ± 10.7 pp, low confidence
155GPT-4.1 mini56.5%estimated ± 10.7 pp, low confidence
156GPT-4.1 nano56.5%estimated ± 10.7 pp, low confidence
157GPT-4o56.5%estimated ± 10.7 pp, low confidence
158GPT-4o mini56.5%estimated ± 10.7 pp, low confidence
159Granite 4.2 3B56.5%estimated ± 10.7 pp, low confidence
160K-Exaone56.5%estimated ± 10.7 pp, low confidence
161LFM2.5-2.6B56.5%estimated ± 10.7 pp, low confidence
162Ling 2.6 Flash56.5%estimated ± 10.7 pp, low confidence
163Llama 3.3 70B56.5%estimated ± 10.7 pp, low confidence
164Mercury 2.556.5%estimated ± 10.7 pp, low confidence
165Mistral Small 3.156.5%estimated ± 10.7 pp, low confidence
166Nemotron 3 Nano Omni 30B A3B56.5%estimated ± 10.7 pp, low confidence
167North Mini Code56.5%estimated ± 10.7 pp, low confidence
168Solar Pro 356.5%estimated ± 10.7 pp, low confidence
169Ultravox v0.6 Llama 3.3 70B56.5%estimated ± 10.7 pp, low confidence
170GPT-5.4 nano56.1%measured
171Kimi K2.655.9%measured
172Claude Sonnet 4.555.8%estimated ± 8.6 pp, medium confidence
173Nemotron 3.5 Bolt 74B-A9B BF1655.6%measured
174Ornith-1.0-9B55.0%estimated ± 13.5 pp, low confidence
175Ornith-1.5-9B54.2%measured
176MiMo-V2-Pro54.2%estimated ± 12.9 pp, low confidence
177Agents-A1-4B54.0%estimated ± 10.3 pp, low confidence
178GPT-5.1-Codex53.9%estimated ± 8.6 pp, medium confidence
179GPT-5.2-Codex53.6%estimated ± 8.6 pp, medium confidence
180Nemotron 3.5 Bolt 74B-A9B52.2%measured
181Qwen3.5-35B-A3B49.4%estimated ± 10.3 pp, low confidence
182GLM-4.748.8%estimated ± 10.3 pp, low confidence
183Sarvam 105B48.8%estimated ± 10.3 pp, low confidence
184Sarvam 30B48.8%estimated ± 10.3 pp, low confidence
185Qwen3.6 Plus48.2%measured
186Claude 4.1 Opus47.8%estimated ± 8.6 pp, medium confidence
187Grok Build 0.147.0%estimated ± 14.6 pp, low confidence
188LLaDA2.2-flash46.2%measured
189Qwen3.5 397B46.1%measured
190LLaDA2.2-mini46.0%estimated ± 13.5 pp, low confidence
191LongCat-Flash-Lite-Sparse45.6%measured
192Grok 4.1 Fast44.7%estimated ± 14.6 pp, low confidence
193GLM-5-Turbo43.9%estimated ± 13.5 pp, low confidence
194Qwen3.5 Plus42.4%estimated ± 8.6 pp, medium confidence
195Claude Opus 4.542.3%measured
196Claude 4 Sonnet42.2%estimated ± 8.6 pp, medium confidence
197Step 3.7 Flash42.0%estimated ± 9.1 pp, medium confidence
198DeepSeek V3.240.9%estimated ± 13.3 pp, low confidence
199GLM-5V-Turbo40.9%estimated ± 13.5 pp, low confidence
200Qwen3 Max40.3%estimated ± 14.6 pp, low confidence
201Pokee-Isaac 28B40.3%estimated ± 14.2 pp, low confidence
202Grok 438.7%estimated ± 14.6 pp, low confidence
203Grok 4.138.7%estimated ± 12.9 pp, low confidence
204Claude Haiku 4.538.1%estimated ± 8.6 pp, medium confidence
205LFM2.5-24B-A2B35.9%estimated ± 14.2 pp, low confidence
206LFM2.5-8B-A1B35.9%estimated ± 14.2 pp, low confidence
207Gemini 3.1 Flash-Lite34.3%estimated ± 14.6 pp, low confidence
208Grok 4.2034.2%estimated ± 14.6 pp, low confidence
209LFM3-1.7B32.6%measured
210GLM-531.1%measured
211Kimi K2.529.5%measured
212Gemini 3 Flash29.2%estimated ± 8.6 pp, medium confidence
213Gemini 3 Pro29.2%estimated ± 8.6 pp, medium confidence
214MiMo-V2-Omni27.9%estimated ± 13.5 pp, low confidence
215GPT-523.0%estimated ± 8.6 pp, low confidence
216GPT-4.121.2%estimated ± 14.6 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General