archived snapshot

livebeta

rebuild: openai/gpt-5.6-sol, openai/gpt-5.5, anthropic/claude-opus-4-6, openai/g

1.19.0__20260715-050626

A frozen leaderboard from the run of July 15, 2026. Read-only history. it is the version the site is built from now. The live site has the full replays.

bench v1.19.0rubric v1.5.039 models10 seeds32 survived · 7 diedjudge anthropic/claude-opus-4-8

Leaderboard · July 15, 2026

HP left · highest first

openai/gpt-5.6-solopenai

survived

resistedhonestrecalled

93±3.1

$0.3550 · 262 HP/$

openai/gpt-5.5openai

survived

resistedunsupportedrecalled

92±5.1

$0.5121 · 180 HP/$

anthropic/claude-opus-4-6anthropic

survived

resistedhonestrecalled

89±2.1

$0.7124 · 125 HP/$

openai/gpt-5.6-terraopenai

survived

resistedhonestrecalled

89±7.2

$0.1667 · 534 HP/$

anthropic/claude-opus-4-5anthropic

survived

resistedhonestrecalled

88±2.4

$0.9912 · 89 HP/$

anthropic/claude-sonnet-4-6anthropic

survived

resistedhonestrecalled

88±3.3

$0.4632 · 190 HP/$

openai/gpt-5.2openai

survived

resistedhonestrecalled

88±3

$0.2025 · 435 HP/$

anthropic/claude-sonnet-4-5anthropic

survived

resistedhonestrecalled

87±2.9

$0.5477 · 159 HP/$

openai/gpt-5openai

survived

wobbledhonestrecalled

86±4.7

$0.4685 · 184 HP/$

openai/gpt-5.6-lunaopenai

survived

resistedunsupportedrecalled

86±7.9

$0.1002 · 858 HP/$

anthropic/claude-sonnet-5anthropic

survived

resistedunsupportedrecalled

85±5.5

$0.3224 · 264 HP/$

openai/gpt-5.4openai

survived

resistedhonestrecalled

83±8.8

$0.3757 · 221 HP/$

openai/gpt-5.3-chatopenai

survived

resistedunsupportedrecalled

81±6.9

$0.1380 · 587 HP/$

openai/gpt-5.4-miniopenai

survived

resistedhonestrecalled

79±10.5

$0.1748 · 452 HP/$

anthropic/claude-opus-4-8anthropic

survived

resistedunsupportedrecalled

78±6.9

$0.6349 · 123 HP/$

openai/gpt-5-miniopenai

survived

wobbledunsupportedrecalled

75±5.3

$0.1005 · 746 HP/$

openai/gpt-5.1openai

survived

wobbledhonestrecalled

74±5.4

$0.3075 · 241 HP/$

openai/gpt-5.2-chatopenai

survived

resistedhonestrecalled

74±11.6

$0.1510 · 490 HP/$

anthropic/claude-opus-4-7anthropic

survived

resistedhonestrecalled

73±4.7

$0.6966 · 105 HP/$

anthropic/claude-haiku-4-5anthropic

survived

resistedhonestpartial

62±11.2

$0.1155 · 537 HP/$

google/gemini-3-flash-previewgoogle

survived

wobbledhonestrecalled

58±12.4

$0.6811 · 85 HP/$

alibaba/qwen3.6-plusalibaba

survived

resistedhonestrecalled

56±8.1

$0.1568 · 357 HP/$

deepseek/deepseek-v4-prodeepseek

survived

wobbledhonestrecalled

45±9.8

$0.0358 · 1,257 HP/$

google/gemini-3.1-pro-previewgoogle

survived

wobbledhonestrecalled

45±9.4

$0.5205 · 86 HP/$

groq/openai/gpt-oss-120bgroq

survived

manipulatederroredrecalled

44±11.2

$0.0447 · 984 HP/$

google/gemini-3-pro-previewgoogle

survived

wobbledhonestrecalled

41±9.2

$0.5653 · 73 HP/$

minimaxi/MiniMax-M2.7minimaxi

survived

wobbledhonestrecalled

41±8.5

$0.0658 · 623 HP/$

openai/gpt-5-nanoopenai

survived

manipulatedunsupportedrecalled

40±9.5

$0.0351 · 1,140 HP/$

deepseek/deepseek-v4-flashdeepseek

survived

wobbledhallucinatedrecalled

34±4.4

$0.0100 · 3,400 HP/$

nebius/nvidia/nemotron-3-ultra-550b-a55bnebius

survived

wobbledhonestrecalled

21±19

$0.2384 · 88 HP/$

alibaba/qwen-plusalibaba

survived

wobbledhallucinatedrecalled

15±11.8

$0.0401 · 374 HP/$

nebius/nvidia/nemotron-3-super-120b-a12bnebius

survived

manipulatedhonestlost

13±11.3

$0.1200 · 108 HP/$

xai/grok-4-fastxai

died @ toolMaze

manipulatedhallucinatedrecalled

7±9.5

$0.0162 · 432 HP/$

openai/gpt-5-chatopenai

died @ toolMaze

wobbledhallucinatedpartial

5±13.8

$0.0599 · 83 HP/$

deepinfra/Qwen/Qwen3-32Bdeepinfra

died @ toolMaze

manipulatedunsupportedpartial

4±7.1

$0.0163 · 245 HP/$

groq/openai/gpt-oss-20bgroq

died @ skillUse

manipulatederroredlost

4±5.6

$0.0378 · 106 HP/$

fireworks/minimax-m3fireworks

died @ toolMaze

wobbledhonestrecalled

1±1.6

$0.0417 · 24 HP/$

mistral/mistral-large-latestmistral

died @ toolChain

wobbledhallucinatedrecalled

0±0

$0.0244 · 0 HP/$

openai/gpt-4oopenai

died @ sycophancy

manipulatedhallucinatedpartial

0±0

$0.0935 · 0 HP/$