Best LLMs for Knowledge in 2026

Top AI models ranked by knowledge benchmarks including MMLU, GPQA, SuperGPQA, and OpenBookQA.

GPT-5.4
OpenAIProprietary1M

96

avg

Gemini 3.1 Pro
GoogleProprietary1M

96

avg

Claude Opus 4.6
AnthropicProprietary1M

96

avg

4
GPT-5.3 Codex
OpenAIProprietary400K

96

avg

5
Grok 4.1
xAIProprietary128K

96

avg

6
GPT-5.2
OpenAIProprietary400K

96

avg

7
GPT-5.2-Codex
OpenAIProprietary400K

96

avg

8
Gemini 3 Pro Deep Think
GoogleProprietary2M

96

avg

9
Claude Sonnet 4.6
AnthropicProprietary1M

96

avg

10
Claude Opus 4.5
AnthropicProprietary200K

96

avg

11
Gemini 3 Pro
GoogleProprietary2M

96

avg

12
GPT-5.1-Codex-Max
OpenAIProprietary400K

95

avg

13
GPT-5.1
OpenAIProprietary400K

94

avg

14
GLM-5 (Reasoning)
Zhipu AIOpen Weight200K

93

avg

15
Claude Sonnet 4.5
AnthropicProprietary1M

92

avg

16
Grok 4.1 Fast
xAIProprietary2M

91

avg

17
GPT-5 (high)
OpenAIProprietary128K

90

avg

18
o1-preview
OpenAIProprietary200K

89

avg

19
Kimi K2.5 (Reasoning)
Moonshot AIOpen Weight128K

89

avg

20
GPT-5 (medium)
OpenAIProprietary128K

88

avg

21
Qwen3.5 397B (Reasoning)
AlibabaOpen Weight128K

88

avg

22
o3-pro
OpenAIProprietary200K

87.3

avg

23
o3
OpenAIProprietary200K

85.3

avg

24
GPT-5 mini
OpenAIProprietary128K

85

avg

25
GLM-5
Zhipu AIOpen Weight200K

85

avg

26
Grok 4
xAIProprietary128K

84.8

avg

27
DeepSeek V3.2 (Thinking)
DeepSeekOpen Weight128K

84

avg

28
GLM-4.7
Zhipu AIOpen Weight200K

83

avg

29
Qwen2.5-1M
AlibabaOpen Weight1M

81.8

avg

30
DeepSeek V3.2
DeepSeekOpen Weight128K

81.8

avg

31
Gemini 2.5 Pro
GoogleProprietary2M

81.5

avg

32
Qwen2.5-72B
AlibabaOpen Weight128K

80.8

avg

33
Qwen3.5 397B
AlibabaOpen Weight128K

80.8

avg

34
o4-mini (high)
OpenAIProprietary200K

80.5

avg

35
DeepSeek Coder 2.0
DeepSeekOpen Weight128K

77.8

avg

36
DeepSeekMath V2
DeepSeekOpen Weight128K

77.8

avg

37
DeepSeek LLM 2.0
DeepSeekOpen Weight128K

76.8

avg

38
MiMo-V2-Flash
XiaomiOpen Weight128K

76.8

avg

39
Kimi K2.5
Moonshot AIOpen Weight128K

74.8

avg

40
Claude 4.1 Opus
AnthropicProprietary200K

74.5

avg

41
Mistral Large 3
MistralOpen Weight128K

73.8

avg

42
Nemotron 3 Ultra 500B
NVIDIAOpen Weight32K

71.8

avg

43
Claude 4 Sonnet
AnthropicProprietary200K

71.5

avg

44
MiniMax M2.5
MiniMaxProprietary128K

70.8

avg

45
Llama 3.1 405B
MetaOpen Weight128K

68.5

avg

46
Gemini 3 Flash
GoogleProprietary1M

67.8

avg

47
Mistral Large 2
MistralProprietary128K

66.5

avg

48
Claude Haiku 4.5
AnthropicProprietary200K

65.8

avg

49
GPT-4o
OpenAIProprietary128K

64.5

avg

50
GLM-4.7-Flash
Zhipu AIOpen Weight200K

63.8

avg

51
Claude 3.5 Sonnet
AnthropicProprietary200K

63.5

avg

52
Mistral 8x7B
MistralOpen Weight32K

62.8

avg

53
Nemotron 3 Super 100B
NVIDIAOpen Weight32K

62.8

avg

54
Gemini 1.5 Pro
GoogleProprietary2M

62.5

avg

55
Grok Code Fast 1
xAIProprietary256K

61.8

avg

56
Gemini 3.1 Flash-Lite
GoogleProprietary1M

60.8

avg

57
Gemini 1.0 Pro
GoogleProprietary32K

60.5

avg

58
Claude 3 Opus
AnthropicProprietary200K

59.5

avg

59
GPT-4 Turbo
OpenAIProprietary128K

58.5

avg

60
Llama 3 70B
MetaOpen Weight128K

56.5

avg

61
Nemotron 3 Nano 30B
NVIDIAOpen Weight32K

54.8

avg

62
Claude 3 Haiku
AnthropicProprietary200K

54.5

avg

63
Nemotron-4 15B
NVIDIAOpen Weight32K

51.8

avg

64
Moonshot v1
Moonshot AIProprietary128K

50.8

avg

65
Z-1
ZProprietary128K

49.8

avg

66
GPT-OSS 120B
OpenAIOpen Weight128K

48.8

avg

67
Gemini 2.5 Flash
GoogleProprietary1M

47.8

avg

68
Nemotron Ultra 253B
NVIDIAOpen Weight32K

46.8

avg

69
Llama 4 Behemoth
MetaOpen Weight32K

45.8

avg

70
Llama 4 Scout
MetaOpen Weight32K

44.8

avg

71
Llama 4 Maverick
MetaOpen Weight32K

43.8

avg

72
Gemma 3 27B
GoogleOpen Weight32K

42.8

avg

73
DeepSeek-R1
DeepSeekOpen Weight128K

41.8

avg

74
Qwen2.5-VL-32B
AlibabaOpen Weight32K

40.8

avg

75
Grok 3 [Beta]
xAIProprietary128K

39.8

avg

76
Nova Pro
Nova AIProprietary128K

38.8

avg

77
Qwen3 235B 2507 (Reasoning)
AlibabaOpen Weight128K

37.8

avg

78
Qwen3 235B 2507
AlibabaOpen Weight128K

36.8

avg

79
Claude 4.1 Opus Thinking
AnthropicProprietary200K

35.8

avg

80
GLM-4.5
TsinghuaProprietary128K

34.8

avg

81
MiniMax M1 80k
MiniMaxProprietary80K

33.8

avg

82
GLM-4.5-Air
TsinghuaProprietary128K

32.8

avg

83
DeepSeek V3.1 (Reasoning)
DeepSeekOpen Weight128K

31.8

avg

84
DeepSeek V3.1
DeepSeekOpen Weight128K

30.8

avg

85
Kimi K2
Moonshot AIProprietary128K

29.8

avg

86
GPT-OSS 20B
OpenAIOpen Weight128K

28.8

avg

87
Mistral 7B v0.3
MistralOpen Weight32K

27.8

avg

88
Mistral 8x7B v0.2
MistralOpen Weight32K

26.8

avg

Key Takeaways

  • The top model is GPT-5.4 by OpenAI with a score of 96.
  • The best open-weight model in this ranking is GLM-5 (Reasoning) at position #14.
  • 88 models are included in this ranking.