livebench.ai · benchmark source

LiveBench

A contamination-aware suite that refreshes questions over time. Category and subtask views live on the source site.

Open livebench.ai1 board on this publisher

Fresh evaluation

LiveBench overall

A benchmark family designed to refresh questions and reduce contamination from training data.

Why it matters

Fresh tasks can provide a less stale view of reasoning and language performance.

Use it when

You want a current external signal alongside static benchmark suites.

The limitation

It still measures selected tasks and cannot predict your full workflow or tool integration.

Leaderboard snapshot

LiveBench overall results in a readable view

Use the tabs for the primary score and other published fields from this board. The source link remains authoritative for the live table.

The current public leaderboard does not expose an exact question-set release identifier, so these scores must not be aligned with results from another capture or release. 58 default-view rows captured 2026-10-01. Top: Claude Fable 5.1 Max Effort (83.4), Claude 5.5 Opus Thinking Max Effort (83.2), Claude Fable 5 Max Effort (83).

Leader among verified rows

Claude Fable 5.1 Max Effort

83.4%

Rows shown

Overall score

Source / snapshot date

2026-10-01

25 days old · not a live API feed

Open the live source before trusting rank order — especially when this snapshot predates recent model releases.

Score profile

Overall score by published row

Higher is better in this view

Claude Fable 5.1 Max Effort

83.4%

Claude 5.5 Opus Thinking Max Effort

83.2%

Claude Fable 5 Max Effort

83%

GPT-6 Astra Max Effort

82.2%

GPT-6.1 Sol Max Effort

81.6%

Muse Spark 1.3 xHigh Effort

81.6%

DeepSeek V4.1 Flash Max Effort

81.1%

GPT-5.6 Sol Max Effort

81%

GPT-5.5 Thinking xHigh Effort

80.2%

Claude 5 Opus Thinking Max Effort

80.1%

GPT-6 Sol Max Effort

79.3%

Kimi K3

79.2%

Gemini 3.7 Flash High

78.8%

Qwen 3.8 Max

78.5%

GPT-5.4 Thinking xHigh Effort

78%

Grok 4.6 xHigh

78%

Muse Spark 1.2 xHigh Effort

78%

GPT-5.6 Terra Max Effort

77.9%

Claude Sonnet 5.5 xHigh Effort

77.8%

DeepSeek V4 Pro 0813

77.4%

Grok 4.7 xHigh

77.4%

Gemini 3.1 Pro Preview High

77%

DeepSeek V4 Flash Vision Exp

76.8%

Claude 4.7 Opus Thinking xHigh Effort

76.5%

Claude 4.8 Opus Thinking Max Effort

76.2%

Qwen 3.8 Flash Next

76.2%

GLM-5.3

76.1%

Claude Sonnet 5 xHigh Effort

76%

Gemini 3.8 Flash High

75.8%

Grok 4.5

75.8%

Muse Spark 1.1 xHigh Effort

75.3%

Qwen3.8 27B

75.3%

Gemini 3.5 Flash High

74.6%

GPT-5.2 High

74.6%

Claude 4.6 Opus Thinking High Effort

74.5%

DeepSeek V4 Flash 0731

74.2%

GPT-5.2 Codex

74%

Gemini 3.6 Flash High

73.6%

GPT-5.6 Luna Max Effort

73.6%

GLM-5.2

73.2%

Qwen 3.7 Max

73.1%

Claude 4.6 Sonnet Thinking Medium Effort

73%

Claude 4.5 Opus Thinking High Effort

72.6%

GPT-6 Luna Max Effort

72%

Inkling xHigh Effort

71.9%

GLM-5.3 Flash

71.6%

Kimi K2.6 Thinking

70.5%

GPT-5.4 Nano xHigh

69.6%

ox-alpha-max

69.2%

Qwen 3.6 Plus

68.9%

Kimi K2.7 Code

68.4%

Grok Build 0.1

67.8%

Nemotron 3 Ultra 550B A55B

67.4%

Minimax M3

67.3%

GPT-5.4 Mini xHigh

66.4%

Qwen 3.6 27B

64%

Gemini 3.5 Flash-Lite High

63.9%

Grok 4.3

62.3%

Benchmark leaderboard snapshot showing rank, score, published details, and notes
Rank	Model or system	Overall score	Cost / successful task	Other details	Note
#1	Claude Fable 5.1 Max EffortAnthropic	83.4%	$1.212	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#2	Claude 5.5 Opus Thinking Max EffortAnthropic	83.2%	$0.799	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#3	Claude Fable 5 Max EffortAnthropic	83%	$1.439	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#4	GPT-6 Astra Max EffortOpenAI	82.2%	$0.736	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#5	GPT-6.1 Sol Max EffortOpenAI	81.6%	$0.142	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#5	Muse Spark 1.3 xHigh EffortMeta	81.6%	$0.219	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#7	DeepSeek V4.1 Flash Max EffortDeepSeek	81.1%	$0.029	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#8	GPT-5.6 Sol Max EffortOpenAI	81%	$0.515	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#9	GPT-5.5 Thinking xHigh EffortOpenAI	80.2%	$0.435	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#10	Claude 5 Opus Thinking Max EffortAnthropic	80.1%	$0.699	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#11	GPT-6 Sol Max EffortOpenAI	79.3%	$0.268	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#12	Kimi K3Moonshot AI	79.2%	$0.348	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#13	Gemini 3.7 Flash HighGoogle	78.8%	$0.157	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#14	Qwen 3.8 MaxQwen	78.5%	$0.275	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#15	GPT-5.4 Thinking xHigh EffortOpenAI	78%	$0.387	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#15	Grok 4.6 xHighSpaceXAI	78%	$0.207	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#15	Muse Spark 1.2 xHigh EffortMeta	78%	$0.375	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#18	GPT-5.6 Terra Max EffortOpenAI	77.9%	$0.352	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#19	Claude Sonnet 5.5 xHigh EffortAnthropic	77.8%	$0.141	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#20	DeepSeek V4 Pro 0813DeepSeek	77.4%	$0.044	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#20	Grok 4.7 xHighSpaceXAI	77.4%	$0.718	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#22	Gemini 3.1 Pro Preview HighGoogle	77%	$0.286	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#23	DeepSeek V4 Flash Vision ExpDeepSeek	76.8%	$0.051	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#24	Claude 4.7 Opus Thinking xHigh EffortAnthropic	76.5%	$0.528	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#25	Claude 4.8 Opus Thinking Max EffortAnthropic	76.2%	$0.983	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#25	Qwen 3.8 Flash NextQwen	76.2%	$0.042	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#27	GLM-5.3Z.ai	76.1%	$0.450	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#28	Claude Sonnet 5 xHigh EffortAnthropic	76%	$0.505	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#29	Gemini 3.8 Flash HighGoogle	75.8%	$0.307	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#29	Grok 4.5SpaceXAI	75.8%	$0.131	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#31	Muse Spark 1.1 xHigh EffortMeta	75.3%	$0.198	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#31	Qwen3.8 27BQwen	75.3%	$0.094	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#33	Gemini 3.5 Flash HighGoogle	74.6%	$0.249	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#33	GPT-5.2 HighOpenAI	74.6%	$0.234	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#35	Claude 4.6 Opus Thinking High EffortAnthropic	74.5%	$0.404	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#36	DeepSeek V4 Flash 0731DeepSeek	74.2%	$0.060	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#37	GPT-5.2 CodexOpenAI	74%	$0.187	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#38	Gemini 3.6 Flash HighGoogle	73.6%	$0.235	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#38	GPT-5.6 Luna Max EffortOpenAI	73.6%	$0.169	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#40	GLM-5.2Z.ai	73.2%	$0.225	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#41	Qwen 3.7 MaxQwen	73.1%	$0.182	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#42	Claude 4.6 Sonnet Thinking Medium EffortAnthropic	73%	$0.306	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#43	Claude 4.5 Opus Thinking High EffortAnthropic	72.6%	$0.610	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#44	GPT-6 Luna Max EffortOpenAI	72%	$0.026	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#45	Inkling xHigh EffortThinking Machines	71.9%	$0.310	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#46	GLM-5.3 FlashZ.ai	71.6%	$0.031	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#47	Kimi K2.6 ThinkingMoonshot AI	70.5%	$0.169	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#48	GPT-5.4 Nano xHighOpenAI	69.6%	$0.091	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#49	ox-alpha-maxNot listed	69.2%	$0.000	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#50	Qwen 3.6 PlusQwen	68.9%	$0.227	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#51	Kimi K2.7 CodeMoonshot AI	68.4%	$0.100	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#52	Grok Build 0.1SpaceXAI	67.8%	$0.024	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#53	Nemotron 3 Ultra 550B A55BNVIDIA	67.4%	$0.371	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#54	Minimax M3MiniMax	67.3%	$0.060	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#55	GPT-5.4 Mini xHighOpenAI	66.4%	$0.334	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#56	Qwen 3.6 27BQwen	64%	$0.202	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#57	Gemini 3.5 Flash-Lite HighGoogle	63.9%	$0.069	—	— Source captured 2026-10-01 · Publisher/run date not disclosed
#58	Grok 4.3SpaceXAI	62.3%	$0.061	—	— Source captured 2026-10-01 · Publisher/run date not disclosed

LiveBench overall table (58 models) captured 2026-10-01 from the live leaderboard's default view: overall and cost per successful task as the site displays them, finetunes hidden and effort variants collapsed to each model's best row. The source displays seven category averages but no exact question-set version in this capture. This capture date is not the evaluation/question-set date. LiveBench refreshes questions over time; scores without a shared exact question-set version are unaligned. Open the source for category/subtask detail.

Open LiveBench leaderboard

Evidence in this catalog

Where LiveBench overall fits

Closest verified examples we currently carry. A missing score is not a zero.

Model examples

Claude Fable 5.1 Max Effort83.4%
Published source observation captured 2026-10-01. Source-specific result; no independent rerun or universal capability verdict.
Claude 5.5 Opus Thinking Max Effort83.2%
Published source observation captured 2026-10-01. Source-specific result; no independent rerun or universal capability verdict.
Claude Fable 5 Max Effort83%
Published source observation captured 2026-10-01. Source-specific result; no independent rerun or universal capability verdict.

Harness examples

No directly comparable harness score is verified here yet.

Open this board on LiveBench