Agents · Benchmark profile

JobBench

An occupational agent benchmark for professional workflows that workers say they most want delegated to AI.

Reference2026Active21 models

Data verified 21 Jul 2026 · Methodology 1.6.0

Benchmark score on JobBench

Leaderboard · 21 models

Muse Spark 1.1 leads at 54.7%, followed by Kimi K3 (52.9%) and Claude Opus 4.7 (Adaptive) (45.9%).

Display onlyAgents2026

#1Closed

Kimi K3

Moonshot AI

kimi-k3

52.9%

Overall #4 · Context 1M tokens

#3Closed

Claude Opus 4.7 (Adaptive)

Anthropic

claude-opus-4-7-max

45.9%

Overall — · Context 1M tokens

Visual analysis

See the field.

Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.

ViewModelsColours

MetaMoonshot AIAnthropicOpenAIAlibaba CloudGoogle

JobBench leaderboard

20 ranked models · higher is better

View accessible chart data

Accessible data for JobBench leaderboard
Model	Rank	Provider	Score
Muse Spark 1.1	#1	Meta	54.7%
Kimi K3	#2	Moonshot AI	52.9%
Claude Opus 4.7 (Adaptive)	#3	Anthropic	45.9%
GPT-5.5	#4	OpenAI	42.7%
GPT-5.4	#5	OpenAI	38.9%
Claude Sonnet 4.6	#6	Anthropic	36.9%
Claude Opus 4.6	#7	Anthropic	36.7%
GPT-5.2	#8	OpenAI	34.3%
GPT-5.3-Codex	#9	OpenAI	33.7%
Claude Opus 4.5	#10	Anthropic	32.3%
Claude Sonnet 4.5	#11	Anthropic	27.7%
GPT-5.1-Codex	#12	OpenAI	26.2%
GPT-5.2 Codex	#13	OpenAI	26%
Claude 4.1 Opus	#14	Anthropic	21.9%
Qwen3.5 Plus	#15	Alibaba Cloud	18.5%
Claude 4 Sonnet	#16	Anthropic	18.4%
Claude Haiku 4.5	#17	Anthropic	16%
Gemini 3 Flash	#18	Google	11.4%
Gemini 3 Pro	#19	Google	11.4%
Kimi K2.5	#20	Moonshot AI	8.7%

Benchmark score table (21 models)

One best score per model · higher is better

JobBench leaderboard ranked by score
Rank	Model	Provider	License	Evidence use	Score
#1	Muse Spark 1.1 muse-spark-1-1	Meta	closed	Reference only	54.7%
#2	Kimi K3 kimi-k3	Moonshot AI	closed	Reference only	52.9%
#3	Claude Opus 4.7 (Adaptive) claude-opus-4-7-max	Anthropic	closed	Estimated reference	45.9%
#4	GPT-5.5 gpt-5.5	OpenAI	closed	Estimated reference	42.7%
#5	GPT-5.4 gpt-5.4	OpenAI	closed	Reference only	38.9%
#6	Claude Sonnet 4.6 claude-sonnet-4-6	Anthropic	closed	Reference only	36.9%
#7	Claude Opus 4.6 claude-opus-4-6	Anthropic	closed	Reference only	36.7%
#8	GPT-5.2 gpt-5.2	OpenAI	closed	Estimated reference	34.3%
#9	GPT-5.3-Codex gpt-5.3-codex	OpenAI	closed	Reference only	33.7%
#10	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	Reference only	32.3%
#11	Claude Sonnet 4.5 claude-sonnet-4-5	Anthropic	closed	Estimated reference	27.7%
#12	GPT-5.1-Codex gpt-5-1-codex	OpenAI	closed	Estimated reference	26.2%
#13	GPT-5.2 Codex gpt-5.2-codex	OpenAI	closed	Reference only	26%
#14	Claude 4.1 Opus claude-4-1-opus	Anthropic	closed	Reference only	21.9%
#15	Qwen3.5 Plus qwen3-5-plus	Alibaba Cloud	closed	Estimated reference	18.5%
#16	Claude 4 Sonnet claude-4-sonnet	Anthropic	closed	Reference only	18.4%
#17	Claude Haiku 4.5 claude-haiku-4-5-20251001	Anthropic	closed	Estimated reference	16%
#18	Gemini 3 Flash gemini-3-flash	Google	closed	Reference only	11.4%
#19	Gemini 3 Pro gemini-3-pro	Google	closed	Reference only	11.4%
#20	Kimi K2.5 kimi-k2.5	Moonshot AI	—	Reference only	8.7%
#21	GPT-5 (high) gpt-5-high	OpenAI	closed	Estimated reference	8.5%

The top of this snapshot is led by Muse Spark 1.1 at 54.7%; third place is 8.8 points behind. The top-10 spread is 22.4 points.

Compare top models

Muse Spark 1.1 vs Kimi K3 Kimi K3 vs Claude Opus 4.7 (Adaptive)Claude Opus 4.7 (Adaptive) vs GPT-5.5 GPT-5.5 vs GPT-5.4

About JobBench

Definition and scoring

Organisation: Yuetai Li, Yichen Feng, Zhangchen Xu, Zixian Ma, Kaiyuan Zheng, Fengqing Jiang, Xinghua Sun, Rulin Shao, Zichen Chen, Yue Huang, Xinyang Han, Brian Lee, Kayla Xu, Shenglai Zeng, Hang Hua, Xiangliang Zhang, Basel Alomair, Ranjay Krishna, Luke Zettlemoyer, Pang Wei Koh, Bhaskar Ramasubramanian, Luyao Niu, Xiang Yue, Radha Poovendran
Category: Agents
Version: 2026
Direction: higher is better
Ranking use: Reference
Contamination risk: Unknown

An occupational agent benchmark for professional workflows that workers say they most want delegated to AI. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.

Open benchmark source ↗

FAQ

Quick answers

What does JobBench measure?

An occupational agent benchmark for professional workflows that workers say they most want delegated to AI.

Which model scores highest on JobBench?

Muse Spark 1.1 by Meta currently leads with 54.7%.

How many models are evaluated on JobBench?

21 models in the LuminaBench cohort have a qualifying score on this benchmark.

Does this affect overall Lumina rank?

No. This benchmark is display-only and does not enter the overall Lumina composite.

Terminal-Bench

2.1 · 155 results · Weighted 1.6%

Berkeley Function-Calling Leaderboard

V3 · 2 results · Not Weighted

τ-bench

Rolling · 37 results · Weighted 1.6%

GAIA

Rolling · 13 results · Weighted 1.6%

BrowseComp

Rolling · 56 results · Weighted 1.6%

OSWorld

Rolling · 13 results · Weighted 1.6%

Agents · Benchmark profile

JobBench

An occupational agent benchmark for professional workflows that workers say they most want delegated to AI.

Reference2026Active21 models

Data verified 21 Jul 2026 · Methodology 1.6.0

Benchmark score on JobBench

Leaderboard · 21 models

Muse Spark 1.1 leads at 54.7%, followed by Kimi K3 (52.9%) and Claude Opus 4.7 (Adaptive) (45.9%).

Display onlyAgents2026

#1Closed

Muse Spark 1.1

Kimi K3

Moonshot AI

kimi-k3

52.9%

Overall #4 · Context 1M tokens

#3Closed

Claude Opus 4.7 (Adaptive)

Anthropic

claude-opus-4-7-max

45.9%

Overall — · Context 1M tokens

Visual analysis

See the field.

Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.

ViewModelsColours

MetaMoonshot AIAnthropicOpenAIAlibaba CloudGoogle

JobBench leaderboard

20 ranked models · higher is better

View accessible chart data

Accessible data for JobBench leaderboard
Model	Rank	Provider	Score
Muse Spark 1.1	#1	Meta	54.7%
Kimi K3	#2	Moonshot AI	52.9%
Claude Opus 4.7 (Adaptive)	#3	Anthropic	45.9%
GPT-5.5	#4	OpenAI	42.7%
GPT-5.4	#5	OpenAI	38.9%
Claude Sonnet 4.6	#6	Anthropic	36.9%
Claude Opus 4.6	#7	Anthropic	36.7%
GPT-5.2	#8	OpenAI	34.3%
GPT-5.3-Codex	#9	OpenAI	33.7%
Claude Opus 4.5	#10	Anthropic	32.3%
Claude Sonnet 4.5	#11	Anthropic	27.7%
GPT-5.1-Codex	#12	OpenAI	26.2%
GPT-5.2 Codex	#13	OpenAI	26%
Claude 4.1 Opus	#14	Anthropic	21.9%
Qwen3.5 Plus	#15	Alibaba Cloud	18.5%
Claude 4 Sonnet	#16	Anthropic	18.4%
Claude Haiku 4.5	#17	Anthropic	16%
Gemini 3 Flash	#18	Google	11.4%
Gemini 3 Pro	#19	Google	11.4%
Kimi K2.5	#20	Moonshot AI	8.7%

Benchmark score table (21 models)

One best score per model · higher is better

JobBench leaderboard ranked by score
Rank	Model	Provider	License	Evidence use	Score
#1	Muse Spark 1.1 muse-spark-1-1	Meta	closed	Reference only	54.7%
#2	Kimi K3 kimi-k3	Moonshot AI	closed	Reference only	52.9%
#3	Claude Opus 4.7 (Adaptive) claude-opus-4-7-max	Anthropic	closed	Estimated reference	45.9%
#4	GPT-5.5 gpt-5.5	OpenAI	closed	Estimated reference	42.7%
#5	GPT-5.4 gpt-5.4	OpenAI	closed	Reference only	38.9%
#6	Claude Sonnet 4.6 claude-sonnet-4-6	Anthropic	closed	Reference only	36.9%
#7	Claude Opus 4.6 claude-opus-4-6	Anthropic	closed	Reference only	36.7%
#8	GPT-5.2 gpt-5.2	OpenAI	closed	Estimated reference	34.3%
#9	GPT-5.3-Codex gpt-5.3-codex	OpenAI	closed	Reference only	33.7%
#10	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	Reference only	32.3%
#11	Claude Sonnet 4.5 claude-sonnet-4-5	Anthropic	closed	Estimated reference	27.7%
#12	GPT-5.1-Codex gpt-5-1-codex	OpenAI	closed	Estimated reference	26.2%
#13	GPT-5.2 Codex gpt-5.2-codex	OpenAI	closed	Reference only	26%
#14	Claude 4.1 Opus claude-4-1-opus	Anthropic	closed	Reference only	21.9%
#15	Qwen3.5 Plus qwen3-5-plus	Alibaba Cloud	closed	Estimated reference	18.5%
#16	Claude 4 Sonnet claude-4-sonnet	Anthropic	closed	Reference only	18.4%
#17	Claude Haiku 4.5 claude-haiku-4-5-20251001	Anthropic	closed	Estimated reference	16%
#18	Gemini 3 Flash gemini-3-flash	Google	closed	Reference only	11.4%
#19	Gemini 3 Pro gemini-3-pro	Google	closed	Reference only	11.4%
#20	Kimi K2.5 kimi-k2.5	Moonshot AI	—	Reference only	8.7%
#21	GPT-5 (high) gpt-5-high	OpenAI	closed	Estimated reference	8.5%

The top of this snapshot is led by Muse Spark 1.1 at 54.7%; third place is 8.8 points behind. The top-10 spread is 22.4 points.

Compare top models

Muse Spark 1.1 vs Kimi K3 Kimi K3 vs Claude Opus 4.7 (Adaptive)Claude Opus 4.7 (Adaptive) vs GPT-5.5 GPT-5.5 vs GPT-5.4

About JobBench

Definition and scoring

Organisation: Yuetai Li, Yichen Feng, Zhangchen Xu, Zixian Ma, Kaiyuan Zheng, Fengqing Jiang, Xinghua Sun, Rulin Shao, Zichen Chen, Yue Huang, Xinyang Han, Brian Lee, Kayla Xu, Shenglai Zeng, Hang Hua, Xiangliang Zhang, Basel Alomair, Ranjay Krishna, Luke Zettlemoyer, Pang Wei Koh, Bhaskar Ramasubramanian, Luyao Niu, Xiang Yue, Radha Poovendran
Category: Agents
Version: 2026
Direction: higher is better
Ranking use: Reference
Contamination risk: Unknown

Open benchmark source ↗

FAQ

Quick answers

What does JobBench measure?

An occupational agent benchmark for professional workflows that workers say they most want delegated to AI.

Which model scores highest on JobBench?

Muse Spark 1.1 by Meta currently leads with 54.7%.

How many models are evaluated on JobBench?

21 models in the LuminaBench cohort have a qualifying score on this benchmark.

Does this affect overall Lumina rank?

No. This benchmark is display-only and does not enter the overall Lumina composite.

Terminal-Bench

2.1 · 155 results · Weighted 1.6%

Berkeley Function-Calling Leaderboard

V3 · 2 results · Not Weighted

τ-bench

Rolling · 37 results · Weighted 1.6%

GAIA

Rolling · 13 results · Weighted 1.6%

BrowseComp

Rolling · 56 results · Weighted 1.6%

OSWorld

Rolling · 13 results · Weighted 1.6%

Search LUMINA BENCH

Navigation

JobBench

Leaderboard · 21 models

Muse Spark 1.1

Kimi K3

Claude Opus 4.7 (Adaptive)

See the field.

JobBench leaderboard

Crop JobBench leaderboard

Benchmark score table (21 models)

Crop JobBench benchmark table

Compare top models

Definition and scoring

Quick answers

What does JobBench measure?

Which model scores highest on JobBench?

How many models are evaluated on JobBench?

Does this affect overall Lumina rank?

JobBench

Leaderboard · 21 models

Muse Spark 1.1

Kimi K3

Claude Opus 4.7 (Adaptive)

See the field.

JobBench leaderboard

Crop JobBench leaderboard

Benchmark score table (21 models)

Crop JobBench benchmark table

Compare top models

Definition and scoring

Quick answers

What does JobBench measure?

Which model scores highest on JobBench?

How many models are evaluated on JobBench?

Does this affect overall Lumina rank?

Leaderboard · 21 models

Muse Spark 1.1

Kimi K3

Claude Opus 4.7 (Adaptive)

See the field.

Benchmark score table (21 models)

Compare top models

Definition and scoring

Quick answers

What does JobBench measure?

Which model scores highest on JobBench?

How many models are evaluated on JobBench?

Does this affect overall Lumina rank?

More Agents benchmarks

Leaderboard · 21 models

Muse Spark 1.1

Kimi K3

Claude Opus 4.7 (Adaptive)

See the field.

Benchmark score table (21 models)

Compare top models

Definition and scoring

Quick answers

What does JobBench measure?

Which model scores highest on JobBench?

How many models are evaluated on JobBench?

Does this affect overall Lumina rank?

More Agents benchmarks