Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

LiveBench (Language)

Evaluates language comprehension through tasks including Connections word puzzles, typo removal, and movie synopsis unscrambling from recent sources.
Source:

Model Performance

#1
Gemini 3.1 Pro Preview
85.4%
#2
Claude Opus 4.6 (Thinking)
83.3%
#3
Claude Opus 4.5 (Thinking)
81.3%
#4
GPT-5
80.8%
#5
GPT-5 (Thinking)
80.7%
#6
Claude Opus 4.5
78.7%
#7
Claude 4.1 Opus
76.8%
#8
GPT-5.1 Codex Max
76.5%
#9
Claude Sonnet 4.5 (Thinking)
76.5%
#10
Claude 4.0 Opus
76.1%
#11
Claude Sonnet 4.6
76.1%
#12
Claude Sonnet 4.5
76.0%
#13
OpenAI o3 (High Effort)
76.0%
#14
GPT-5 Mini (Thinking)
75.5%
#15
Gemini 2.5 Pro (Thinking)
75.5%
#16
Gemini 2.5 Pro
74.5%
#17
Claude 4.0 Opus (Thinking)
73.7%
#18
OpenAI o3 (Medium Effort)
73.5%
#19
Claude 4.0 Sonnet (Thinking)
72.9%
#20
Claude 4.1 Opus (Thinking)
72.8%
#21
Claude 4.0 Sonnet
71.0%
#22
DeepSeek V3.2 (Thinking)
70.4%
#23
Qwen 3 (Thinking)
69.5%
#24
GPT-5 Mini
68.8%
#25
Claude 3.7 Sonnet (Thinking)
68.3%
#26
Kimi K2
66.7%
#27
Claude Haiku 4.5 (Thinking)
66.5%
#28
Qwen 3 Max Preview
66.1%
#29
OpenAI o4 Mini (High Effort)
66.0%
#30
Gemini 2.5 Flash (Thinking)
65.3%
#31
DeepSeek R1
64.8%
#32
GPT-4.5
64.8%
#33
DeepSeek V3.2
64.2%
#34
Claude 3.7 Sonnet
63.2%
#35
OpenAI o4 Mini (Medium Effort)
62.4%
#36
DeepSeek V3
59.8%
#37
Grok 3 Mini
59.1%
#38
Claude Haiku 4.5
57.0%
#39
Gemini 2.5 Flash
57.0%
#40
Qwen 3
55.5%
#41
GPT-4.1
54.5%
#42
Claude 3.5 Sonnet
54.5%
#43
GPT-5 Nano
51.0%
#44
GPT-5.2
50.0%
#45
Llama 4 Maverick 17B
49.6%
#46
GPT-4o
49.4%
#47
GPT OSS 120B
48.6%
#48
GPT-5 Nano (Thinking)
46.8%
#49
Mistral Medium 3.1
44.7%
#50
Mistral Large
40.5%
#51
Claude 3.5 Haiku
39.7%
#52
Cohere Command A
36.7%
#53
Cohere Command R+
30.9%
#54
Phi-4
30.7%