Vision & documents

CharXiv w/o tools leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on CharXiv w/o tools is 88.9% by Claude Mythos 5. 108 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Mythos 5	88.9%	measured
2	Gemini 3.1 Pro	88.6%	estimated ± 1.8 pp, low confidence
3	Qwen3.8 Max	88.4%	measured
4	Gemini 3.5 Flash	88.3%	estimated ± 1.8 pp, low confidence
5	GPT-5.6 Sol	87.7%	estimated ± 1.8 pp, low confidence
6	Claude Opus 5.5	87.2%	estimated ± 2.4 pp, medium confidence
7	GPT-6 Astra	86.7%	estimated ± 2.4 pp, medium confidence
8	GPT-6.1 Sol	86.2%	estimated ± 2.4 pp, medium confidence
9	Gemini 3.8 Flash	86.2%	measured
10	GPT-5.4	85.7%	estimated ± 1.8 pp, medium confidence
11	GPT-5.5	85.7%	estimated ± 1.8 pp, medium confidence
12	Claude Opus 5	85.5%	estimated ± 2.4 pp, high confidence
13	GPT-5.6 Terra	85.2%	estimated ± 1.8 pp, medium confidence
14	Muse Spark	84.9%	estimated ± 1.8 pp, medium confidence
15	Kimi K3	84.8%	measured
16	Qwen3.8-Flash-Next	84.6%	measured
17	Gemini 3.6 Flash	84.6%	estimated ± 2.4 pp, high confidence
18	Gemini 3.7 Flash	84.5%	measured
19	GPT-6 Sol	84.4%	estimated ± 2.4 pp, high confidence
20	Qwen3.8 Max Preview	84.4%	estimated ± 2.4 pp, high confidence
21	Seed 2.1 Pro	83.8%	estimated ± 1.7 pp, medium confidence
22	Qwen3.7 Plus	83.7%	estimated ± 1.7 pp, medium confidence
23	Seed 2.1 Turbo	83.7%	estimated ± 1.7 pp, medium confidence
24	Qwen3.5 397B	83.7%	estimated ± 1.7 pp, medium confidence
25	Qwen3.6 Plus	83.7%	estimated ± 1.7 pp, medium confidence
26	Claude Opus 4.5	83.7%	estimated ± 1.7 pp, low confidence
27	Gemini 3 Pro	83.7%	estimated ± 1.7 pp, low confidence
28	Gemma 4 12B	83.7%	estimated ± 1.7 pp, low confidence
29	GPT-5.2	83.7%	estimated ± 1.7 pp, low confidence
30	Kimi K2.6	83.7%	estimated ± 1.7 pp, low confidence
31	Qwen3.5-122B-A10B	83.7%	estimated ± 1.7 pp, low confidence
32	Qwen3.5-27B	83.7%	estimated ± 1.7 pp, low confidence
33	Qwen3.5-35B-A3B	83.7%	estimated ± 1.7 pp, low confidence
34	Qwen3.8-27B	83.7%	measured
35	Qwen3.8-Omni-Flash	83.5%	measured
36	dots3-note Preview	83.1%	measured
37	Grok 4.5	82.9%	estimated ± 2.4 pp, high confidence
38	Kimi K2.5	82.8%	estimated ± 1.8 pp, medium confidence
39	Kimi K2.5 (Reasoning)	82.8%	estimated ± 1.8 pp, medium confidence
40	GPT-5.6 Luna	82.7%	estimated ± 1.8 pp, medium confidence
41	GPT-6 Luna	82.5%	estimated ± 2.4 pp, high confidence
42	Grok 4.3	82.4%	estimated ± 1.8 pp, medium confidence
43	MiniMax M3	82.4%	estimated ± 1.8 pp, medium confidence
44	Pareto 26.9	82.3%	estimated ± 1.8 pp, medium confidence
45	Apodex 1.1	82.2%	estimated ± 2.4 pp, high confidence
46	Apodex 1.1 Mini	82.2%	estimated ± 2.4 pp, high confidence
47	MiMo-V2.5	82.1%	estimated ± 1.8 pp, medium confidence
48	Claude Opus 4.7 (Adaptive)	82.1%	measured
49	Gemini 3.5 Flash-Lite	82.1%	estimated ± 2.4 pp, high confidence
50	Ling 3.0 Flash VL	82.1%	estimated ± 2.4 pp, high confidence
51	Gemini 3 Flash	81.8%	estimated ± 2.4 pp, high confidence
52	GPT-5.3 Codex	81.8%	estimated ± 2.4 pp, high confidence
53	GLM-5.3-Flash	81.5%	estimated ± 2.3 pp, high confidence
54	Claude Opus 4.6	81.5%	estimated ± 1.8 pp, medium confidence
55	Gemma 4 31B	81.1%	estimated ± 1.8 pp, medium confidence
56	DeepSeek V4.1 Flash	80.8%	estimated ± 2.4 pp, high confidence
57	GPT-5.4 mini	80.7%	estimated ± 1.8 pp, medium confidence
58	Muse Spark 1.1	80.5%	estimated ± 2.3 pp, high confidence
59	Claude Opus 4.8	80.5%	measured
60	Claude Opus 4.7	80.5%	estimated ± 2.4 pp, high confidence
61	Mistral Large 4	80.5%	estimated ± 2.4 pp, high confidence
62	GPT-5.2-Codex	80.4%	estimated ± 2.4 pp, high confidence
63	Step 5 Preview	80.1%	estimated ± 1.8 pp, medium confidence
64	GPT-5.1	79.9%	estimated ± 2.4 pp, high confidence
65	Qwen3.6-27B	79.9%	estimated ± 1.8 pp, medium confidence
66	Claude Opus 4.6 (Adaptive)	79.8%	estimated ± 2.4 pp, high confidence
67	Step 3.7 Flash	79.8%	estimated ± 2.4 pp, high confidence
68	Gemini 2.5 Pro	79.5%	estimated ± 2.4 pp, high confidence
69	Qwen3.6-35B-A3B	79.3%	estimated ± 1.8 pp, medium confidence
70	Sakana Fugu-Ultra	79.2%	estimated ± 2.3 pp, high confidence
71	Grok 4.20	79.2%	estimated ± 1.8 pp, medium confidence
72	GPT-5 (medium)	79.1%	estimated ± 2.4 pp, high confidence
73	GPT-5 (high)	79.1%	estimated ± 2.4 pp, high confidence
74	Claude Opus 4.5 Thinking	78.9%	estimated ± 2.4 pp, high confidence
75	Sakana Fugu	78.5%	estimated ± 2.3 pp, high confidence
76	MiMo-V2.6-Flash	78.4%	estimated ± 2.4 pp, medium confidence
77	GLM-5V-Turbo	78.2%	estimated ± 2.4 pp, medium confidence
78	Inkling	78.1%	measured
79	GPT-5.1-Codex	78.0%	estimated ± 2.4 pp, medium confidence
80	GPT-5.1-Codex-Max	78.0%	estimated ± 2.4 pp, medium confidence
81	Muse Glimmer 30B	77.9%	estimated ± 1.8 pp, medium confidence
82	Gemma 4 26B A4B	77.7%	estimated ± 1.8 pp, medium confidence
83	Inkling-Small	77.4%	measured
84	Claude Sonnet 4.6	77.3%	estimated ± 2.3 pp, medium confidence
85	Nemotron 3 Nano Omni 30B A3B	77.3%	estimated ± 2.3 pp, medium confidence
86	Gemini 3.1 Flash-Lite	77.2%	estimated ± 2.3 pp, medium confidence
87	Claude Sonnet 5	77.0%	measured
88	o3	76.4%	estimated ± 2.4 pp, medium confidence
89	MiMo-V2-Omni	76.2%	estimated ± 2.4 pp, medium confidence
90	Grok 4	75.5%	estimated ± 2.4 pp, medium confidence
91	Claude 4.1 Opus Thinking	74.9%	estimated ± 2.4 pp, medium confidence
92	Interfaze Beta	74.8%	estimated ± 1.8 pp, low confidence
93	Gemini 2.5 Flash	73.2%	estimated ± 2.4 pp, medium confidence
94	Mistral Medium 3.5 128B	72.8%	estimated ± 2.4 pp, medium confidence
95	Grok 4.1 Fast (Reasoning)	71.6%	estimated ± 2.4 pp, medium confidence
96	Claude 4 Sonnet	71.0%	estimated ± 2.4 pp, medium confidence
97	Llama 4 Maverick	70.7%	estimated ± 2.4 pp, medium confidence
98	Grok 4 Fast (Reasoning)	70.5%	estimated ± 2.4 pp, medium confidence
99	GPT-4.1	70.1%	estimated ± 2.4 pp, medium confidence
100	GPT-5.4 nano	69.4%	estimated ± 1.8 pp, low confidence
101	Qwen3-Omni-30B-A3B-Thinking	69.3%	estimated ± 2.4 pp, medium confidence
102	GPT-4.1 mini	68.2%	estimated ± 2.4 pp, medium confidence
103	Mistral Small 4	66.7%	estimated ± 2.4 pp, medium confidence
104	Mistral Small 4 (Reasoning)	66.7%	estimated ± 2.4 pp, medium confidence
105	Command A+	66.0%	estimated ± 1.8 pp, low confidence
106	Mistral Large 3	65.8%	estimated ± 2.4 pp, medium confidence
107	Qwen3-Omni-30B-A3B-Instruct	65.7%	estimated ± 2.4 pp, medium confidence
108	Gemini 1.5 Pro	65.3%	estimated ± 2.4 pp, medium confidence
109	Mistral Medium 3	63.7%	estimated ± 2.4 pp, medium confidence
110	Llama 4 Scout	63.6%	estimated ± 2.4 pp, medium confidence
111	Qwen3.5 397B (Reasoning)	63.4%	estimated ± 2.4 pp, medium confidence
112	Gemma 4 E4B	62.3%	estimated ± 2.4 pp, medium confidence
113	Grok 4.1 Fast	59.8%	estimated ± 2.4 pp, medium confidence
114	Gemma 3 27B	59.4%	estimated ± 2.4 pp, medium confidence
115	Gemma 4 E2B	56.4%	estimated ± 2.4 pp, medium confidence
116	Nova Pro	56.1%	estimated ± 2.4 pp, medium confidence
117	GPT-4o mini	53.5%	estimated ± 2.4 pp, medium confidence
118	GPT-4.1 nano	52.2%	estimated ± 2.4 pp, medium confidence
119	Claude 3 Haiku	42.7%	estimated ± 2.4 pp, medium confidence
120	LFM2.5-VL-1.6B-Extract	37.9%	estimated ± 2.4 pp, medium confidence
121	LFM2.5-VL-3B	30.8%	estimated ± 1.8 pp, low confidence
122	Phi-4 Multimodal Instruct	22.7%	estimated ± 2.4 pp, medium confidence

CharXiv w/o tools leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General