Agents · Benchmark profile

τ-bench

A benchmark for tool-using language agents interacting with simulated user and business environments.

Ranking weightedCurrent / rollingActive37 models

Data verified 15 July 2026 · Methodology 1.2.0

Benchmark score on τ-bench

Leaderboard · 37 models

GLM-5.2 leads at 99.1%, followed by Claude Fable 5 (98.5%) and GLM-5-Turbo (98.5%).

Ranking · 1.6%AgentsCurrent

#1Open

GLM-5.2

Z.AI

glm-5-2

99.1%

Overall #11 · Context 1M tokens

#2Closed

Claude Fable 5

Anthropic

claude-fable-5

98.5%

Overall #2 · Context 1M tokens

#3Closed

GLM-5-Turbo

Z.AI

glm-5-turbo

98.5%

Overall #28 · Context 200K tokens

Visual analysis

See the field, not just the table.

Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.

ModelsColours

Z.AIAnthropicxAIAlibaba CloudMoonshot AIGoogleXiaomiOpenAIMetaMiniMax

Leaderboard bars

20 ranked models · higher is better

View accessible chart data

Accessible data for Leaderboard bars
Model	Rank	Provider	Score
GLM-5.2	#1	Z.AI	99.1%
Claude Fable 5	#2	Anthropic	98.5%
GLM-5-Turbo	#3	Z.AI	98.5%
GLM-5V-Turbo	#4	Z.AI	98.5%
GLM-5	#5	Z.AI	98.2%
GLM-5.1	#6	Z.AI	97.7%
Grok 4.3	#7	xAI	97.7%
Qwen3.6 Plus	#8	Alibaba Cloud	97.7%
Kimi K2.5	#9	Moonshot AI	95.9%
Gemini 3.1 Pro Preview	#10	Google	95.6%
Gemini 3.5 Flash	#11	Google	95.3%
MiMo-V2-Pro	#12	Xiaomi	95%
Qwen3.7-Max	#13	Alibaba Cloud	94.7%
Claude Opus 4.8	#14	Anthropic	94.4%
MiMo-V2.5-Pro	#15	Xiaomi	94.2%
GPT-5.5	#16	OpenAI	93.9%
Qwen3.7-Plus	#17	Alibaba Cloud	93%
Muse Spark	#18	Meta	91.5%
MiMo-V2-Omni	#19	Xiaomi	91.2%
MiniMax M3	#20	MiniMax	88.9%

Score curve

Shows how quickly scores separate down the field

View accessible chart data

Accessible data for Score curve
Model	Rank	Provider	Score
GLM-5.2	#1	Z.AI	99.1%
Claude Fable 5	#2	Anthropic	98.5%
GLM-5-Turbo	#3	Z.AI	98.5%
GLM-5V-Turbo	#4	Z.AI	98.5%
GLM-5	#5	Z.AI	98.2%
GLM-5.1	#6	Z.AI	97.7%
Grok 4.3	#7	xAI	97.7%
Qwen3.6 Plus	#8	Alibaba Cloud	97.7%
Kimi K2.5	#9	Moonshot AI	95.9%
Gemini 3.1 Pro Preview	#10	Google	95.6%
Gemini 3.5 Flash	#11	Google	95.3%
MiMo-V2-Pro	#12	Xiaomi	95%
Qwen3.7-Max	#13	Alibaba Cloud	94.7%
Claude Opus 4.8	#14	Anthropic	94.4%
MiMo-V2.5-Pro	#15	Xiaomi	94.2%
GPT-5.5	#16	OpenAI	93.9%
Qwen3.7-Plus	#17	Alibaba Cloud	93%
Muse Spark	#18	Meta	91.5%
MiMo-V2-Omni	#19	Xiaomi	91.2%
MiniMax M3	#20	MiniMax	88.9%

Provider profile

Average within the visible cohort · descriptive only

View accessible chart data

Accessible data for Provider profile
Provider	Models	Average score
Z.AI	5	98.4%
xAI	1	97.7%
Anthropic	2	96.5%
Moonshot AI	1	95.9%
Google	2	95.4%
Alibaba Cloud	3	95.1%
OpenAI	1	93.9%
Xiaomi	3	93.5%
Meta	1	91.5%
MiniMax	1	88.9%

Benchmark score table (37 models)

One best score per model · higher is better

τ-bench leaderboard ranked by score
Rank	Model	Provider	License	Score
#1	GLM-5.2 glm-5-2	Z.AI	open	99.1%
#2	Claude Fable 5 claude-fable-5	Anthropic	closed	98.5%
#3	GLM-5-Turbo glm-5-turbo	Z.AI	closed	98.5%
#4	GLM-5V-Turbo glm-5v-turbo	Z.AI	closed	98.5%
#5	GLM-5 glm-5	Z.AI	open	98.2%
#6	GLM-5.1 glm-5-1	Z.AI	open	97.7%
#7	Grok 4.3 grok-4.3	xAI	closed	97.7%
#8	Qwen3.6 Plus qwen3-6-plus	Alibaba Cloud	closed	97.7%
#9	Kimi K2.5 kimi-k2.5	Moonshot AI	—	95.9%
#10	Gemini 3.1 Pro Preview gemini-3.1-pro-preview	Google	closed	95.6%
#11	Gemini 3.5 Flash gemini-3.5-flash	Google	closed	95.3%
#12	MiMo-V2-Pro mimo-v2-pro	Xiaomi	closed	95%
#13	Qwen3.7-Max qwen3.7-max	Alibaba Cloud	—	94.7%
#14	Claude Opus 4.8 claude-opus-4-8	Anthropic	closed	94.4%
#15	MiMo-V2.5-Pro mimo-v2-5-pro	Xiaomi	closed	94.2%
#16	GPT-5.5 gpt-5.5	OpenAI	closed	93.9%
#17	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	93%
#18	Muse Spark muse-spark	Meta	closed	91.5%
#19	MiMo-V2-Omni mimo-v2-omni	Xiaomi	closed	91.2%
#20	MiniMax M3 MiniMax-M3	MiniMax	—	88.9%
#21	Gemini 3 Pro gemini-3-pro	Google	closed	87.1%
#22	GPT-5.4 gpt-5.4	OpenAI	closed	87.1%
#23	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	86.3%
#24	GPT-5.6 Terra gpt-5.6-terra	OpenAI	closed	86.3%
#25	GPT-5.3-Codex gpt-5.3-codex	OpenAI	closed	86%
#26	GPT-5.6 Sol gpt-5.6-sol	OpenAI	closed	85.1%
#27	Claude Opus 4.6 claude-opus-4-6	Anthropic	closed	84.8%
#28	MiniMax M2.7 minimax-m2-7	MiniMax	open	84.8%
#29	GPT-5.1 gpt.5.1	OpenAI	closed	81.9%
#30	Claude Sonnet 4.6 claude-sonnet-4-6	Anthropic	closed	79.5%
#31	GPT-5.4 nano gpt.5.4.nano	OpenAI	closed	76%
#32	Claude Opus 4.7 claude-opus-4-7	Anthropic	closed	74%
#33	Gemma 4 31B gemma-4-31b	Google	open	59.9%
#34	Gemini 3 Flash gemini-3-flash	Google	closed	43.3%
#35	Mistral Small 4 mistral-small-2603	Mistral AI	open	41.2%
#36	Gemini 3.1 Flash-Lite gemini-3.1-flash-lite	Google	closed	31.3%
#37	Mistral Large 3 mistral-large-2512	Mistral AI	open	24.6%

The top of this snapshot is led by GLM-5.2 at 99.1%; third place is 0.60 points behind. The top-10 spread is 3.5 points.

Compare top models

GLM-5.2 vs Claude Fable 5 Claude Fable 5 vs GLM-5-Turbo GLM-5-Turbo vs GLM-5V-Turbo GLM-5V-Turbo vs GLM-5

About τ-bench

Definition and scoring

Organisation: Sierra Research
Category: Agents
Version: Not evaluated
Direction: higher is better
Ranking use: Weighted 1.6%
Contamination risk: Unknown

A benchmark for tool-using language agents interacting with simulated user and business environments. 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 τ-bench measure?

A benchmark for tool-using language agents interacting with simulated user and business environments.

Which model scores highest on τ-bench?

GLM-5.2 by Z.AI currently leads with 99.1%.

How many models are evaluated on τ-bench?

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

Does this affect overall Lumina rank?

Yes. This family is ranking-weighted at 1.6% of its capability category. GLM-5.2 is overall #11.

Terminal-Bench

2.1 · 80 results · Weighted 1.6%

Berkeley Function-Calling Leaderboard

V3 · 2 results · Not Weighted

GAIA

Rolling · 13 results · Weighted 1.6%

BrowseComp

Rolling · 24 results · Weighted 1.6%

OSWorld

Rolling · 11 results · Weighted 1.6%

WebArena

Rolling · 2 results · Not Weighted

Agents · Benchmark profile

τ-bench

A benchmark for tool-using language agents interacting with simulated user and business environments.

Ranking weightedCurrent / rollingActive37 models

Data verified 15 July 2026 · Methodology 1.2.0

Benchmark score on τ-bench

Leaderboard · 37 models

GLM-5.2 leads at 99.1%, followed by Claude Fable 5 (98.5%) and GLM-5-Turbo (98.5%).

Ranking · 1.6%AgentsCurrent

#1Open

GLM-5.2

Z.AI

glm-5-2

99.1%

Overall #11 · Context 1M tokens

#2Closed

Claude Fable 5

Anthropic

claude-fable-5

98.5%

Overall #2 · Context 1M tokens

#3Closed

GLM-5-Turbo

Z.AI

glm-5-turbo

98.5%

Overall #28 · Context 200K tokens

Visual analysis

See the field, not just the table.

Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.

ModelsColours

Z.AIAnthropicxAIAlibaba CloudMoonshot AIGoogleXiaomiOpenAIMetaMiniMax

Leaderboard bars

20 ranked models · higher is better

View accessible chart data

Accessible data for Leaderboard bars
Model	Rank	Provider	Score
GLM-5.2	#1	Z.AI	99.1%
Claude Fable 5	#2	Anthropic	98.5%
GLM-5-Turbo	#3	Z.AI	98.5%
GLM-5V-Turbo	#4	Z.AI	98.5%
GLM-5	#5	Z.AI	98.2%
GLM-5.1	#6	Z.AI	97.7%
Grok 4.3	#7	xAI	97.7%
Qwen3.6 Plus	#8	Alibaba Cloud	97.7%
Kimi K2.5	#9	Moonshot AI	95.9%
Gemini 3.1 Pro Preview	#10	Google	95.6%
Gemini 3.5 Flash	#11	Google	95.3%
MiMo-V2-Pro	#12	Xiaomi	95%
Qwen3.7-Max	#13	Alibaba Cloud	94.7%
Claude Opus 4.8	#14	Anthropic	94.4%
MiMo-V2.5-Pro	#15	Xiaomi	94.2%
GPT-5.5	#16	OpenAI	93.9%
Qwen3.7-Plus	#17	Alibaba Cloud	93%
Muse Spark	#18	Meta	91.5%
MiMo-V2-Omni	#19	Xiaomi	91.2%
MiniMax M3	#20	MiniMax	88.9%

Score curve

Shows how quickly scores separate down the field

View accessible chart data

Accessible data for Score curve
Model	Rank	Provider	Score
GLM-5.2	#1	Z.AI	99.1%
Claude Fable 5	#2	Anthropic	98.5%
GLM-5-Turbo	#3	Z.AI	98.5%
GLM-5V-Turbo	#4	Z.AI	98.5%
GLM-5	#5	Z.AI	98.2%
GLM-5.1	#6	Z.AI	97.7%
Grok 4.3	#7	xAI	97.7%
Qwen3.6 Plus	#8	Alibaba Cloud	97.7%
Kimi K2.5	#9	Moonshot AI	95.9%
Gemini 3.1 Pro Preview	#10	Google	95.6%
Gemini 3.5 Flash	#11	Google	95.3%
MiMo-V2-Pro	#12	Xiaomi	95%
Qwen3.7-Max	#13	Alibaba Cloud	94.7%
Claude Opus 4.8	#14	Anthropic	94.4%
MiMo-V2.5-Pro	#15	Xiaomi	94.2%
GPT-5.5	#16	OpenAI	93.9%
Qwen3.7-Plus	#17	Alibaba Cloud	93%
Muse Spark	#18	Meta	91.5%
MiMo-V2-Omni	#19	Xiaomi	91.2%
MiniMax M3	#20	MiniMax	88.9%

Provider profile

Average within the visible cohort · descriptive only

View accessible chart data

Accessible data for Provider profile
Provider	Models	Average score
Z.AI	5	98.4%
xAI	1	97.7%
Anthropic	2	96.5%
Moonshot AI	1	95.9%
Google	2	95.4%
Alibaba Cloud	3	95.1%
OpenAI	1	93.9%
Xiaomi	3	93.5%
Meta	1	91.5%
MiniMax	1	88.9%

Benchmark score table (37 models)

One best score per model · higher is better

τ-bench leaderboard ranked by score
Rank	Model	Provider	License	Score
#1	GLM-5.2 glm-5-2	Z.AI	open	99.1%
#2	Claude Fable 5 claude-fable-5	Anthropic	closed	98.5%
#3	GLM-5-Turbo glm-5-turbo	Z.AI	closed	98.5%
#4	GLM-5V-Turbo glm-5v-turbo	Z.AI	closed	98.5%
#5	GLM-5 glm-5	Z.AI	open	98.2%
#6	GLM-5.1 glm-5-1	Z.AI	open	97.7%
#7	Grok 4.3 grok-4.3	xAI	closed	97.7%
#8	Qwen3.6 Plus qwen3-6-plus	Alibaba Cloud	closed	97.7%
#9	Kimi K2.5 kimi-k2.5	Moonshot AI	—	95.9%
#10	Gemini 3.1 Pro Preview gemini-3.1-pro-preview	Google	closed	95.6%
#11	Gemini 3.5 Flash gemini-3.5-flash	Google	closed	95.3%
#12	MiMo-V2-Pro mimo-v2-pro	Xiaomi	closed	95%
#13	Qwen3.7-Max qwen3.7-max	Alibaba Cloud	—	94.7%
#14	Claude Opus 4.8 claude-opus-4-8	Anthropic	closed	94.4%
#15	MiMo-V2.5-Pro mimo-v2-5-pro	Xiaomi	closed	94.2%
#16	GPT-5.5 gpt-5.5	OpenAI	closed	93.9%
#17	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	93%
#18	Muse Spark muse-spark	Meta	closed	91.5%
#19	MiMo-V2-Omni mimo-v2-omni	Xiaomi	closed	91.2%
#20	MiniMax M3 MiniMax-M3	MiniMax	—	88.9%
#21	Gemini 3 Pro gemini-3-pro	Google	closed	87.1%
#22	GPT-5.4 gpt-5.4	OpenAI	closed	87.1%
#23	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	86.3%
#24	GPT-5.6 Terra gpt-5.6-terra	OpenAI	closed	86.3%
#25	GPT-5.3-Codex gpt-5.3-codex	OpenAI	closed	86%
#26	GPT-5.6 Sol gpt-5.6-sol	OpenAI	closed	85.1%
#27	Claude Opus 4.6 claude-opus-4-6	Anthropic	closed	84.8%
#28	MiniMax M2.7 minimax-m2-7	MiniMax	open	84.8%
#29	GPT-5.1 gpt.5.1	OpenAI	closed	81.9%
#30	Claude Sonnet 4.6 claude-sonnet-4-6	Anthropic	closed	79.5%
#31	GPT-5.4 nano gpt.5.4.nano	OpenAI	closed	76%
#32	Claude Opus 4.7 claude-opus-4-7	Anthropic	closed	74%
#33	Gemma 4 31B gemma-4-31b	Google	open	59.9%
#34	Gemini 3 Flash gemini-3-flash	Google	closed	43.3%
#35	Mistral Small 4 mistral-small-2603	Mistral AI	open	41.2%
#36	Gemini 3.1 Flash-Lite gemini-3.1-flash-lite	Google	closed	31.3%
#37	Mistral Large 3 mistral-large-2512	Mistral AI	open	24.6%

The top of this snapshot is led by GLM-5.2 at 99.1%; third place is 0.60 points behind. The top-10 spread is 3.5 points.

Compare top models

GLM-5.2 vs Claude Fable 5 Claude Fable 5 vs GLM-5-Turbo GLM-5-Turbo vs GLM-5V-Turbo GLM-5V-Turbo vs GLM-5

About τ-bench

Definition and scoring

Organisation: Sierra Research
Category: Agents
Version: Not evaluated
Direction: higher is better
Ranking use: Weighted 1.6%
Contamination risk: Unknown

Open benchmark source ↗

FAQ

Quick answers

What does τ-bench measure?

A benchmark for tool-using language agents interacting with simulated user and business environments.

Which model scores highest on τ-bench?

GLM-5.2 by Z.AI currently leads with 99.1%.

How many models are evaluated on τ-bench?

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

Does this affect overall Lumina rank?

Yes. This family is ranking-weighted at 1.6% of its capability category. GLM-5.2 is overall #11.

Terminal-Bench

2.1 · 80 results · Weighted 1.6%

Berkeley Function-Calling Leaderboard

V3 · 2 results · Not Weighted

GAIA

Rolling · 13 results · Weighted 1.6%

BrowseComp

Rolling · 24 results · Weighted 1.6%

OSWorld

Rolling · 11 results · Weighted 1.6%

WebArena

Rolling · 2 results · Not Weighted

Leaderboard · 37 models

GLM-5.2

Claude Fable 5

GLM-5-Turbo

See the field, not just the table.

Benchmark score table (37 models)

Compare top models

Definition and scoring

Quick answers

What does τ-bench measure?

Which model scores highest on τ-bench?

How many models are evaluated on τ-bench?

Does this affect overall Lumina rank?

More Agents benchmarks

Leaderboard · 37 models

GLM-5.2

Claude Fable 5

GLM-5-Turbo

See the field, not just the table.

Benchmark score table (37 models)

Compare top models

Definition and scoring

Quick answers

What does τ-bench measure?

Which model scores highest on τ-bench?

How many models are evaluated on τ-bench?

Does this affect overall Lumina rank?

More Agents benchmarks