benchgap
Vision & documents

ParseBench-1 leaderboard

As of 2026-11-18, the highest measured score on ParseBench-1 is 78.8% by Claude Opus 5.5. 111 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.286.6%estimated ± 4.6 pp, low confidence
2Claude Opus 5.578.8%measured
3GPT-6.2 Astra78.0%estimated ± 6.2 pp, low confidence
4Claude Fable 5.177.8%measured
5Gemini 4 Argon77.5%estimated ± 3.7 pp, high confidence
6Claude Opus 577.5%estimated ± 4.6 pp, medium confidence
7GPT-5.6 Sol74.4%measured
8Gemini 3 Flash74.1%measured
9Hy473.7%measured
10GPT-6.1 Terra73.6%measured
11Gemini 3.5 Flash73.2%measured
12Gemini 4 Hydrogen73.0%estimated ± 5.7 pp, low confidence
13Glimpse72.9%estimated ± 8.6 pp, low confidence
14Hy4 preview72.1%estimated ± 5.4 pp, low confidence
15Kimi K3.1 Pro71.2%estimated ± 1.0 pp, low confidence
16Claude Mythos 571.2%estimated ± 1.0 pp, low confidence
17Qwen3.8 Max71.2%estimated ± 1.0 pp, low confidence
18Gemini 3.8 Flash71.2%measured
19Inkling 2 Small71.1%estimated ± 1.0 pp, medium confidence
20Kimi K370.7%estimated ± 1.0 pp, medium confidence
21Qwen3.8-Flash-Next70.6%estimated ± 1.0 pp, medium confidence
22GLM-5.3-Flash70.6%measured
23Gemini 3.7 Flash70.4%measured
24GPT-6 Astra69.8%measured
25Claude Fable 569.8%measured
26GPT-6.1 Sol69.6%measured
27Claude Sonnet 5.569.2%measured
28Muse Ember69.1%estimated ± 3.7 pp, high confidence
29Gemini 3.6 Flash69.1%measured
30Qwen3.8-27B69.0%measured
31LongCat-2.569.0%estimated ± 8.6 pp, low confidence
32Qwen3.8 Max Preview69.0%estimated ± 6.2 pp, medium confidence
33GPT-6 Sol68.9%measured
34Qwen3.8-Omni-Flash68.3%estimated ± 1.0 pp, medium confidence
35Step 5 Preview68.0%estimated ± 5.4 pp, low confidence
36Gemini 3.1 Pro67.3%measured
37GPT-5.6 Terra67.1%measured
38dots3-note Preview66.9%estimated ± 1.0 pp, medium confidence
39GPT-5.6 Luna66.9%measured
40Seed 2.1 Pro66.8%estimated ± 8.4 pp, medium confidence
41Muse Spark 1.166.6%estimated ± 8.6 pp, low confidence
42Muse Spark Open66.5%estimated ± 6.2 pp, medium confidence
43GPT-5.566.4%measured
44Grok 4.565.8%estimated ± 6.2 pp, medium confidence
45Seed 2.1 Turbo64.5%estimated ± 8.4 pp, medium confidence
46DeepSeek V4.1 Flash 101764.4%measured
47Apodex 1.164.3%estimated ± 6.2 pp, medium confidence
48Apodex 1.1 Mini64.3%estimated ± 6.2 pp, medium confidence
49Sakana Fugu-Ultra64.2%estimated ± 8.6 pp, low confidence
50MiniMax M3.1 Flash64.1%estimated ± 6.2 pp, medium confidence
51GPT-6 Luna64.1%measured
52Ling 3.0 Flash VL64.0%estimated ± 6.2 pp, medium confidence
53GPT-5.263.6%estimated ± 8.4 pp, medium confidence
54GPT-5.3 Codex63.4%estimated ± 6.2 pp, medium confidence
55Claude Opus 4.763.3%measured
56Kimi K2.663.0%estimated ± 5.5 pp, low confidence
57Grok 4.362.9%estimated ± 6.2 pp, medium confidence
58MiniMax M362.8%estimated ± 5.4 pp, low confidence
59Gemma 4 31B62.4%measured
60Claude Opus 4.862.3%measured
61Sakana Fugu62.2%estimated ± 8.6 pp, low confidence
62Muse Spark61.9%estimated ± 5.2 pp, low confidence
63GPT-5.461.7%measured
64Inkling61.6%estimated ± 1.0 pp, medium confidence
65Inkling-Small61.6%estimated ± 1.0 pp, medium confidence
66LFM3-VL-2B61.6%estimated ± 1.0 pp, low confidence
67Pareto 26.961.4%estimated ± 8.4 pp, medium confidence
68MiMo-V2.561.2%estimated ± 8.4 pp, medium confidence
69Claude Sonnet 561.2%measured
70GPT-5.2-Codex60.7%estimated ± 6.2 pp, medium confidence
71GPT-5.159.7%estimated ± 6.2 pp, medium confidence
72Claude Opus 4.6 (Adaptive)59.6%estimated ± 6.2 pp, medium confidence
73Kimi K2.559.6%estimated ± 6.2 pp, medium confidence
74Step 3.7 Flash59.5%estimated ± 6.2 pp, medium confidence
75Gemini 3.5 Flash-Lite59.5%measured
76Qwen3.7 Plus59.3%estimated ± 5.2 pp, low confidence
77Qwen3.5-122B-A10B59.1%estimated ± 6.2 pp, medium confidence
78Qwen3.5-27B59.1%estimated ± 6.2 pp, medium confidence
79GPT-5.4 mini59.0%estimated ± 5.5 pp, low confidence
80Gemini 2.5 Pro59.0%estimated ± 6.2 pp, medium confidence
81Qwen3.6-27B58.7%estimated ± 6.2 pp, medium confidence
82Gemma 4 26B A4B58.5%measured
83GPT-558.2%estimated ± 6.2 pp, medium confidence
84Hy4 flash58.1%estimated ± 8.4 pp, medium confidence
85Claude Opus 4.5 Thinking58.0%estimated ± 6.2 pp, medium confidence
86Muse Glimmer 30B57.5%estimated ± 5.2 pp, low confidence
87Grok 4.2057.3%estimated ± 8.4 pp, medium confidence
88MiMo-V2.6-Flash57.0%estimated ± 6.2 pp, medium confidence
89Nemotron 3.5 Super VL 120B A12B57.0%estimated ± 8.6 pp, low confidence
90GLM-5V-Turbo56.7%estimated ± 6.2 pp, medium confidence
91Gemini 3.1 Flash-Lite56.7%measured
92Qwen3.5-35B-A3B56.6%estimated ± 6.2 pp, medium confidence
93GPT-5.1-Codex56.4%estimated ± 6.2 pp, medium confidence
94GPT-5.1-Codex-Max56.4%estimated ± 6.2 pp, medium confidence
95Gemini 3 Pro56.0%estimated ± 5.2 pp, low confidence
96Claude Haiku 5.556.0%measured
97Nemotron 3.5 Super VL 120B A12B NVFP455.9%estimated ± 8.6 pp, low confidence
98Holo2-235B-A22B54.9%estimated ± 5.2 pp, low confidence
99Claude Sonnet 4.654.4%estimated ± 6.2 pp, medium confidence
100o353.9%estimated ± 6.2 pp, medium confidence
101Claude Opus 4.653.8%measured
102MiMo-V2-Omni53.7%estimated ± 6.2 pp, medium confidence
103Qwen3.6 Plus53.6%estimated ± 5.2 pp, low confidence
104Gemma 4 12B53.5%estimated ± 6.2 pp, medium confidence
105o4-mini53.1%estimated ± 6.2 pp, medium confidence
106Grok 452.7%estimated ± 6.2 pp, medium confidence
107Holo2-30B-A3B52.4%estimated ± 5.2 pp, low confidence
108Qwen3.5 397B52.1%estimated ± 5.2 pp, low confidence
109Interfaze Beta51.7%estimated ± 8.4 pp, medium confidence
110Claude Haiku 4.5 Thinking51.2%measured
111Gemini 2.5 Flash49.9%estimated ± 6.2 pp, medium confidence
112Mistral Medium 3.5 128B49.4%estimated ± 6.2 pp, medium confidence
113Claude Haiku 4.549.3%measured
114Nemotron 3 Nano Omni 30B A3B48.5%measured
115Grok 4.1 Fast (Reasoning)48.3%estimated ± 6.2 pp, medium confidence
116Holo2-8B48.2%estimated ± 5.2 pp, low confidence
117Command A+48.2%estimated ± 6.2 pp, medium confidence
118Claude 4 Sonnet47.7%estimated ± 6.2 pp, medium confidence
119Llama 4 Maverick47.5%estimated ± 6.2 pp, medium confidence
120Grok 4 Fast (Reasoning)47.3%estimated ± 6.2 pp, medium confidence
121Holo2-4B47.2%estimated ± 5.2 pp, low confidence
122GPT-4.146.9%estimated ± 6.2 pp, medium confidence
123Qwen3-Omni-30B-A3B-Thinking46.3%estimated ± 6.2 pp, medium confidence
124Claude 3.7 Sonnet46.3%estimated ± 6.2 pp, medium confidence
125GPT-4.1 mini45.5%estimated ± 6.2 pp, medium confidence
126Mistral Small 444.6%estimated ± 6.2 pp, medium confidence
127Mistral Large 344.2%estimated ± 6.2 pp, medium confidence
128Qwen3-Omni-30B-A3B-Instruct44.1%estimated ± 6.2 pp, medium confidence
129Gemini 1.5 Pro43.9%estimated ± 6.2 pp, medium confidence
130Gemini 2.5 Flash-Lite43.5%estimated ± 6.2 pp, medium confidence
131Mistral Medium 343.2%estimated ± 6.2 pp, medium confidence
132Llama 4 Scout43.2%estimated ± 6.2 pp, medium confidence
133GPT-5.4 nano42.2%measured
134Gemini 1.5 Flash42.0%estimated ± 6.2 pp, low confidence
135Grok 4.1 Fast42.0%estimated ± 6.2 pp, low confidence
136Gemma 3 27B41.9%estimated ± 6.2 pp, low confidence
137Claude 3.5 Haiku41.4%estimated ± 6.2 pp, low confidence
138Gemma 4 E2B41.3%estimated ± 6.2 pp, low confidence
139Nova Pro41.3%estimated ± 6.2 pp, low confidence
140Gemma 4 E4B41.1%measured
141GPT-4o mini40.9%estimated ± 6.2 pp, low confidence
142GPT-4.1 nano40.8%estimated ± 6.2 pp, low confidence
143Qwen3.5-4B40.6%estimated ± 3.7 pp, high confidence
144Amazon Nova 2 Lite40.4%measured
145Claude 3 Haiku40.3%estimated ± 6.2 pp, low confidence
146LFM2.5-VL-1.6B-Extract40.3%estimated ± 6.2 pp, low confidence
147Claude Opus 4.539.9%estimated ± 5.2 pp, low confidence
148Qwen3.6-35B-A3B37.4%measured
149Qwen2.5-VL-32B35.0%estimated ± 8.4 pp, low confidence
150Mistral Small 3.134.9%estimated ± 8.4 pp, low confidence
151LFM2.5-VL-3B32.4%estimated ± 8.4 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General