benchgap
Vision & documents

CharXiv w/o tools leaderboard

As of 2026-11-21, the highest measured score on CharXiv w/o tools is 89.0% by Kimi K3.1 Pro. 113 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.1 Terra90.9%estimated ± 1.2 pp, medium confidence
2Glimpse90.5%estimated ± 2.1 pp, low confidence
3Claude Opus 5.590.2%estimated ± 1.8 pp, low confidence
4Claude Fable 5.189.6%estimated ± 1.8 pp, low confidence
5Kimi K3.1 Pro89.0%measured
6Claude Mythos 588.9%measured
7Gemini 3.1 Pro88.9%estimated ± 1.2 pp, high confidence
8GPT-6.2 Astra88.8%estimated ± 2.5 pp, medium confidence
9Gemini 3.5 Flash88.5%estimated ± 1.2 pp, high confidence
10Qwen3.8 Max88.4%measured
11GPT-5.6 Sol87.7%estimated ± 1.2 pp, high confidence
12Muse Spark Open87.6%estimated ± 2.0 pp, medium confidence
13Gemini 4 Argon87.6%estimated ± 2.5 pp, low confidence
14Gemini 3 Flash87.4%estimated ± 1.8 pp, low confidence
15Muse Spark 1.186.8%estimated ± 2.0 pp, medium confidence
16Gemini 3.8 Flash86.2%measured
17Claude Opus 586.1%estimated ± 2.5 pp, high confidence
18Seed 2.1 Pro86.0%estimated ± 1.2 pp, high confidence
19Inkling 2 Small85.6%measured
20GPT-5.485.5%estimated ± 1.2 pp, high confidence
21GPT-5.585.5%estimated ± 1.2 pp, high confidence
22Gemini 3 Pro85.3%estimated ± 1.2 pp, high confidence
23LongCat-2.585.1%estimated ± 1.9 pp, low confidence
24Qwen3.8 Max Preview85.0%estimated ± 2.5 pp, high confidence
25GPT-5.6 Terra85.0%estimated ± 1.2 pp, high confidence
26Kimi K384.8%measured
27GPT-6 Astra84.8%estimated ± 1.8 pp, medium confidence
28Claude Fable 584.7%estimated ± 1.8 pp, medium confidence
29GPT-6.1 Sol84.6%estimated ± 1.8 pp, medium confidence
30Muse Spark84.6%estimated ± 1.2 pp, high confidence
31Qwen3.8-Flash-Next84.6%measured
32Gemini 3.7 Flash84.5%measured
33GLM-5.3-Flash84.5%estimated ± 1.8 pp, medium confidence
34Claude Sonnet 5.584.4%estimated ± 1.8 pp, medium confidence
35Gemini 3.6 Flash84.3%estimated ± 1.8 pp, medium confidence
36Seed 2.1 Turbo84.3%estimated ± 1.2 pp, high confidence
37GPT-6 Sol84.1%estimated ± 1.8 pp, medium confidence
38Qwen3.8-27B83.7%measured
39GPT-5.283.6%estimated ± 1.2 pp, high confidence
40Grok 4.583.5%estimated ± 2.5 pp, high confidence
41Kimi K2.683.5%estimated ± 1.2 pp, high confidence
42Qwen3.8-Omni-Flash83.5%measured
43dots3-note Preview83.1%measured
44Qwen3.5 397B83.1%estimated ± 1.2 pp, high confidence
45Qwen3.7 Plus83.1%estimated ± 1.2 pp, high confidence
46Qwen3.6 Plus82.9%estimated ± 1.2 pp, high confidence
47Apodex 1.182.8%estimated ± 2.5 pp, high confidence
48Apodex 1.1 Mini82.8%estimated ± 2.5 pp, high confidence
49Ling 3.0 Flash VL82.7%estimated ± 2.5 pp, high confidence
50Hy482.7%estimated ± 1.2 pp, high confidence
51Kimi K2.582.6%estimated ± 1.2 pp, high confidence
52GPT-5.6 Luna82.5%estimated ± 1.2 pp, high confidence
53GPT-5.3 Codex82.4%estimated ± 2.5 pp, high confidence
54Grok 4.382.2%estimated ± 1.2 pp, high confidence
55MiniMax M382.2%estimated ± 1.2 pp, high confidence
56Claude Opus 4.782.1%measured
57Pareto 26.982.1%estimated ± 1.2 pp, high confidence
58MiMo-V2.582.0%estimated ± 1.2 pp, high confidence
59Claude Opus 4.681.4%estimated ± 1.2 pp, high confidence
60DeepSeek V4.1 Flash 101781.2%estimated ± 1.8 pp, medium confidence
61GPT-5.2-Codex81.0%estimated ± 2.5 pp, high confidence
62Gemma 4 31B81.0%estimated ± 1.2 pp, high confidence
63Qwen3.5-122B-A10B81.0%estimated ± 1.2 pp, high confidence
64GPT-6 Luna81.0%estimated ± 1.8 pp, medium confidence
65GPT-5.4 mini80.7%estimated ± 1.2 pp, high confidence
66Inkling 2 Mini80.6%measured
67GPT-5.180.5%estimated ± 2.5 pp, high confidence
68Claude Opus 4.880.5%measured
69Claude Opus 4.6 (Adaptive)80.4%estimated ± 2.5 pp, high confidence
70Step 3.7 Flash80.4%estimated ± 2.5 pp, high confidence
71Step 5 Preview80.2%estimated ± 1.2 pp, high confidence
72Gemini 2.5 Pro80.1%estimated ± 2.5 pp, high confidence
73Hy4 flash80.0%estimated ± 1.2 pp, high confidence
74Qwen3.6-27B80.0%estimated ± 1.2 pp, high confidence
75Sakana Fugu-Ultra79.9%estimated ± 2.4 pp, high confidence
76GPT-579.7%estimated ± 2.5 pp, high confidence
77Qwen3.6-35B-A3B79.6%estimated ± 1.2 pp, high confidence
78Claude Opus 4.5 Thinking79.5%estimated ± 2.5 pp, high confidence
79Grok 4.2079.5%estimated ± 1.2 pp, high confidence
80Qwen3.5-35B-A3B79.4%estimated ± 1.2 pp, high confidence
81Qwen3.5-27B79.3%estimated ± 1.2 pp, high confidence
82Sakana Fugu79.2%estimated ± 2.4 pp, high confidence
83MiMo-V2.6-Flash79.0%estimated ± 2.5 pp, medium confidence
84GLM-5V-Turbo78.8%estimated ± 2.5 pp, medium confidence
85GPT-5.1-Codex78.6%estimated ± 2.5 pp, medium confidence
86GPT-5.1-Codex-Max78.6%estimated ± 2.5 pp, medium confidence
87Holo2-235B-A22B78.5%estimated ± 5.1 pp, low confidence
88Muse Glimmer 30B78.5%estimated ± 1.2 pp, high confidence
89MiniMax M3.1 Flash78.4%estimated ± 2.4 pp, high confidence
90Gemma 4 26B A4B78.3%estimated ± 1.2 pp, high confidence
91Inkling78.1%measured
92Claude Sonnet 4.677.8%estimated ± 2.4 pp, medium confidence
93Nemotron 3 Nano Omni 30B A3B77.8%estimated ± 2.4 pp, medium confidence
94Gemini 3.5 Flash-Lite77.7%estimated ± 1.8 pp, low confidence
95Inkling-Small77.4%measured
96Claude Sonnet 577.0%measured
97o377.0%estimated ± 2.5 pp, medium confidence
98MiMo-V2-Omni76.8%estimated ± 2.5 pp, medium confidence
99Nemotron 3.5 Super VL 120B A12B76.5%estimated ± 1.6 pp, high confidence
100Interfaze Beta76.4%estimated ± 1.2 pp, high confidence
101Holo2-30B-A3B76.4%estimated ± 5.1 pp, low confidence
102Nemotron 3.5 Super VL 120B A12B NVFP476.4%estimated ± 1.6 pp, high confidence
103o4-mini76.3%estimated ± 2.5 pp, medium confidence
104Claude Opus 4.576.1%estimated ± 1.2 pp, high confidence
105Grok 476.1%estimated ± 2.5 pp, medium confidence
106Gemini 3.1 Flash-Lite75.5%estimated ± 1.8 pp, low confidence
107Gemma 4 12B75.3%estimated ± 1.2 pp, high confidence
108Claude Haiku 5.575.1%estimated ± 1.8 pp, low confidence
109Qwen3-VL-235B-A22B Instruct74.8%estimated ± 1.2 pp, high confidence
110GPT-5.4 nano73.9%estimated ± 1.2 pp, high confidence
111GLM-4.6V73.8%estimated ± 1.2 pp, high confidence
112Gemini 2.5 Flash73.8%estimated ± 2.5 pp, medium confidence
113Qwen3.5-4B73.7%estimated ± 2.5 pp, medium confidence
114Mistral Medium 3.5 128B73.3%estimated ± 2.5 pp, medium confidence
115Command A+72.8%estimated ± 1.2 pp, high confidence
116Holo2-8B72.6%estimated ± 5.1 pp, low confidence
117Ternary Bonsai 2 27B72.2%estimated ± 3.8 pp, medium confidence
118Grok 4.1 Fast (Reasoning)72.2%estimated ± 2.5 pp, medium confidence
119GLM-4.6V-Flash72.1%estimated ± 1.2 pp, high confidence
120Holo2-4B71.6%estimated ± 5.1 pp, low confidence
121Claude 4 Sonnet71.5%estimated ± 2.5 pp, medium confidence
122Llama 4 Maverick71.3%estimated ± 2.5 pp, medium confidence
123Claude Haiku 4.5 Thinking71.1%estimated ± 1.8 pp, low confidence
124Grok 4 Fast (Reasoning)71.1%estimated ± 2.5 pp, medium confidence
125MiniMax-VL-0170.8%estimated ± 1.2 pp, high confidence
126Gemma 4 E4B70.7%estimated ± 1.2 pp, high confidence
127Qwen2.5-VL-32B70.4%estimated ± 1.2 pp, high confidence
128Mistral Small 3.170.4%estimated ± 1.2 pp, high confidence
129LFM3-VL-2B70.2%measured
130Gemma 4 E2B70.1%estimated ± 1.2 pp, medium confidence
131LFM2.5-VL-3B69.8%estimated ± 1.2 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General