benchgap
General

SpireClimb leaderboard

As of 2026-11-19, the highest measured score on SpireClimb is 41.3% by GPT-6.2 Astra. 101 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.2 Astra41.3%measured
2GPT-6.2 Sol39.5%estimated ± 2.4 pp, medium confidence
3Gemini 4 Carbon38.2%estimated ± 5.4 pp, low confidence
4GPT-6 Cyber35.5%estimated ± 2.4 pp, medium confidence
5GPT-6 Astra34.0%measured
6Gemini 4 Argon30.0%estimated ± 5.4 pp, low confidence
7GPT-6.1 Sol23.8%measured
8Gemini 3.1 Pro22.4%estimated ± 8.8 pp, medium confidence
9Gemini 3.8 Flash22.4%estimated ± 8.8 pp, medium confidence
10GPT-6.1 Terra22.3%measured
11Claude Fable 5.215.3%measured
12GPT-5.5 Pro13.6%estimated ± 10.9 pp, low confidence
13Gemini 3 Pro12.4%estimated ± 8.8 pp, medium confidence
14Claude Opus 512.2%estimated ± 8.8 pp, medium confidence
15Gemini 3.7 Flash11.7%estimated ± 8.8 pp, medium confidence
16GPT-5.4 Pro11.3%estimated ± 10.9 pp, low confidence
17Kimi K311.0%estimated ± 8.8 pp, medium confidence
18Kimi K3.1 Pro10.7%estimated ± 8.8 pp, medium confidence
19Muse Wonder10.7%estimated ± 8.8 pp, medium confidence
20Claude Fable 59.8%estimated ± 8.8 pp, medium confidence
21Gemini 3.5 Flash9.6%estimated ± 8.8 pp, medium confidence
22MiMo-V2.6-Pro9.3%estimated ± 5.4 pp, low confidence
23GPT-6 Sol9.2%measured
24Claude Opus 4.6 (Adaptive)9.1%estimated ± 8.8 pp, medium confidence
25GPT-5.48.3%estimated ± 8.8 pp, medium confidence
26Claude Opus 4.88.1%estimated ± 8.8 pp, medium confidence
27Gemini 4 Hydrogen7.7%measured
28GPT-5.6 Sol7.1%estimated ± 2.4 pp, medium confidence
29Grok 4.20 Multi-agent7.0%estimated ± 8.8 pp, medium confidence
30DeepSeek V4.1 Flash 10176.9%estimated ± 5.4 pp, low confidence
31Gemini 3.6 Flash6.6%estimated ± 8.8 pp, medium confidence
32Qwen3.8 Max6.2%estimated ± 8.8 pp, medium confidence
33Claude Opus 5.55.8%measured
34Grok 4.1 Fast (Reasoning)5.7%estimated ± 8.8 pp, medium confidence
35Kimi K2.65.6%estimated ± 8.8 pp, medium confidence
36Grok 4.204.9%estimated ± 8.8 pp, medium confidence
37Muse Spark 1.34.8%estimated ± 8.8 pp, medium confidence
38Qwen3.7 Max4.8%estimated ± 8.8 pp, medium confidence
39Muse Spark 1.14.7%estimated ± 8.8 pp, medium confidence
40GPT-5.24.3%estimated ± 8.8 pp, medium confidence
41Hy44.3%estimated ± 8.8 pp, medium confidence
42Gemini 3 Flash4.2%estimated ± 8.8 pp, medium confidence
43Grok 4.63.4%estimated ± 8.8 pp, medium confidence
44Grok 4.53.4%estimated ± 8.8 pp, medium confidence
45Muse Spark 1.23.3%estimated ± 8.8 pp, medium confidence
46GPT-5.6 Terra3.1%estimated ± 8.8 pp, medium confidence
47Muse Spark Open3.1%estimated ± 8.8 pp, medium confidence
48GLM-5.13.0%estimated ± 8.8 pp, medium confidence
49Grok 4.72.9%estimated ± 8.8 pp, medium confidence
50Step 5 Preview2.8%estimated ± 8.8 pp, medium confidence
51Claude Sonnet 52.6%estimated ± 8.8 pp, medium confidence
52Qwen3.7 Plus2.6%estimated ± 8.8 pp, medium confidence
53GLM-5.22.5%estimated ± 8.8 pp, medium confidence
54GLM-5.32.5%estimated ± 8.8 pp, medium confidence
55Qwen 3.6 Max (preview)2.5%estimated ± 8.8 pp, medium confidence
56Gemma 4 31B2.1%estimated ± 8.8 pp, medium confidence
57Claude Fable 5.12.0%measured
58GPT-5.6 Luna2.0%estimated ± 8.8 pp, medium confidence
59Hy4 preview1.9%estimated ± 8.8 pp, low confidence
60Claude Haiku 5.51.7%estimated ± 8.8 pp, low confidence
61MiniMax M31.6%estimated ± 8.8 pp, low confidence
62GPT-5.4 mini1.4%estimated ± 8.8 pp, low confidence
63MiMo-V2.6-Flash1.4%estimated ± 5.4 pp, low confidence
64Gemini 3.5 Flash-Lite1.4%estimated ± 8.8 pp, low confidence
65Qwen3.6 Plus1.3%estimated ± 8.8 pp, low confidence
66Claude Sonnet 5.51.3%measured
67Grok 4.31.1%estimated ± 8.8 pp, low confidence
68GPT-5.51.1%estimated ± 5.4 pp, low confidence
69Qwen3.8-27B1.1%estimated ± 8.8 pp, low confidence
70Claude Opus 4.5 Thinking1.0%estimated ± 8.8 pp, low confidence
71Qwen3.5-122B-A10B1.0%estimated ± 8.8 pp, low confidence
72Claude Opus 4.50.9%estimated ± 8.8 pp, low confidence
73Qwen3.5-27B0.8%estimated ± 8.8 pp, low confidence
74Claude Sonnet 4.60.7%estimated ± 8.8 pp, low confidence
75Qwen3.7 Flash0.7%estimated ± 8.8 pp, low confidence
76Claude Opus 4.60.7%estimated ± 8.8 pp, low confidence
77Qwen3.6-35B-A3B0.6%estimated ± 8.8 pp, low confidence
78Hy30.6%estimated ± 8.8 pp, low confidence
79Step 3.7 Flash0.6%estimated ± 8.8 pp, low confidence
80DeepSeek V4 Flash 07310.2%estimated ± 9.7 pp, medium confidence
81o30.1%estimated ± 9.7 pp, medium confidence
82Claude Opus 4.70.1%estimated ± 9.7 pp, medium confidence
83GPT-50.0%estimated ± 9.7 pp, medium confidence
84Nemotron 3 Ultra0.0%estimated ± 9.7 pp, medium confidence
85Qwen3.5 Flash0.0%estimated ± 9.7 pp, medium confidence
86Qwen3.5 397B0.0%estimated ± 9.7 pp, medium confidence
87Claude Sonnet 4.50.0%estimated ± 9.7 pp, medium confidence
88Mistral Large 40.0%estimated ± 9.7 pp, medium confidence
89Qwen3.5 Plus0.0%estimated ± 9.7 pp, medium confidence
90GPT-5.10.0%estimated ± 9.7 pp, medium confidence
91GPT-4o mini0.0%estimated ± 9.7 pp, medium confidence
92GLM-5.3-Flash0.0%estimated ± 9.7 pp, medium confidence
93GPT-4.1 mini0.0%estimated ± 9.7 pp, medium confidence
94o3-mini0.0%estimated ± 9.7 pp, medium confidence
95Qwen3.6-27B0.0%estimated ± 9.7 pp, medium confidence
96Inkling-Small0.0%estimated ± 9.7 pp, low confidence
97GPT-5.4 nano0.0%estimated ± 9.7 pp, low confidence
98o4-mini0.0%estimated ± 9.7 pp, low confidence
99Qwen3 Max0.0%estimated ± 9.7 pp, low confidence
100GPT-OSS 120B0.0%estimated ± 9.7 pp, low confidence
101Claude 3 Opus0.0%estimated ± 10.9 pp, low confidence
102Claude Haiku 4.50.0%estimated ± 10.9 pp, low confidence
103GLM-4.70.0%estimated ± 10.9 pp, low confidence
104GPT-4.10.0%estimated ± 10.9 pp, low confidence
105GPT-4.1 nano0.0%estimated ± 10.9 pp, low confidence
106GPT-4o0.0%estimated ± 10.9 pp, low confidence
107GPT-6 Luna0.0%measured
108Inkling0.0%estimated ± 10.9 pp, low confidence
109Kimi K2.7 Code0.0%estimated ± 10.9 pp, low confidence
110o10.0%estimated ± 10.9 pp, low confidence
111Seed 2.1 Pro0.0%estimated ± 10.9 pp, low confidence
112Seed 2.1 Turbo0.0%estimated ± 10.9 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General