Agents · Benchmark profile

Terminal-Bench

A benchmark for agents performing practical tasks in reproducible terminal environments.

Ranking weighted2.1Rolling50 models

Data verified 15 July 2026 · Methodology 1.2.0

Benchmark score on Terminal-Bench

Leaderboard · 50 models

GPT-5.6 Sol leads at 91.9%, followed by Kimi K3 (88.3%) and GPT-5.6 Terra (88.0%).

Ranking · 1.6%Agents2.1

#1Closed

GPT-5.6 Sol

OpenAI

gpt-5.6-sol

91.9%

Overall #4 · Context 1.05M tokens

#2Closed

Kimi K3

Moonshot AI

kimi-k3

88.3%

Overall #3 · Context 1M tokens

#3Closed

GPT-5.6 Terra

OpenAI

gpt-5.6-terra

88.0%

Overall #9 · Context 1.05M 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

OpenAIMoonshot AIAnthropicxAIZ.AIMetaGoogleAlibaba CloudXiaomi

Leaderboard bars

20 ranked models · higher is better

View accessible chart data

Accessible data for Leaderboard bars
Model	Rank	Provider	Score
GPT-5.6 Sol	#1	OpenAI	91.9%
Kimi K3	#2	Moonshot AI	88.3%
GPT-5.6 Terra	#3	OpenAI	88.0%
Claude Mythos 5	#4	Anthropic	88%
GPT-5.6 Luna	#5	OpenAI	84.7%
Claude Fable 5	#6	Anthropic	84.6%
Claude Opus 4.8	#7	Anthropic	84.6%
GPT-5.5	#8	OpenAI	84.3%
Grok 4.5	#9	xAI	83.3%
GLM-5.2	#10	Z.AI	81%
Claude Sonnet 5	#11	Anthropic	80.5%
Muse Spark 1.1	#12	Meta	80%
GPT-5.4	#13	OpenAI	78.3%
Gemini 3.5 Flash	#14	Google	76.2%
Qwen3.7-Max	#15	Alibaba Cloud	74.5%
Gemini 3 Pro	#16	Google	73.9%
Gemini 3.1 Pro Preview	#17	Google	73.8%
Qwen3.7-Plus	#18	Alibaba Cloud	70.3%
Claude Opus 4.7	#19	Anthropic	69.4%
MiMo-V2.5-Pro	#20	Xiaomi	68.4%

Score curve

Shows how quickly scores separate down the field

View accessible chart data

Accessible data for Score curve
Model	Rank	Provider	Score
GPT-5.6 Sol	#1	OpenAI	91.9%
Kimi K3	#2	Moonshot AI	88.3%
GPT-5.6 Terra	#3	OpenAI	88.0%
Claude Mythos 5	#4	Anthropic	88%
GPT-5.6 Luna	#5	OpenAI	84.7%
Claude Fable 5	#6	Anthropic	84.6%
Claude Opus 4.8	#7	Anthropic	84.6%
GPT-5.5	#8	OpenAI	84.3%
Grok 4.5	#9	xAI	83.3%
GLM-5.2	#10	Z.AI	81%
Claude Sonnet 5	#11	Anthropic	80.5%
Muse Spark 1.1	#12	Meta	80%
GPT-5.4	#13	OpenAI	78.3%
Gemini 3.5 Flash	#14	Google	76.2%
Qwen3.7-Max	#15	Alibaba Cloud	74.5%
Gemini 3 Pro	#16	Google	73.9%
Gemini 3.1 Pro Preview	#17	Google	73.8%
Qwen3.7-Plus	#18	Alibaba Cloud	70.3%
Claude Opus 4.7	#19	Anthropic	69.4%
MiMo-V2.5-Pro	#20	Xiaomi	68.4%

Provider profile

Average within the visible cohort · descriptive only

View accessible chart data

Accessible data for Provider profile
Provider	Models	Average score
Moonshot AI	1	88.3%
OpenAI	5	85.4%
xAI	1	83.3%
Anthropic	5	81.4%
Z.AI	1	81%
Meta	1	80%
Google	3	74.6%
Alibaba Cloud	2	72.4%
Xiaomi	1	68.4%

Benchmark score table (50 models)

One best score per model · higher is better

Terminal-Bench leaderboard ranked by score
Rank	Model	Provider	License	Score
#1	GPT-5.6 Sol gpt-5.6-sol	OpenAI	closed	91.9%
#2	Kimi K3 kimi-k3	Moonshot AI	closed	88.3%
#3	GPT-5.6 Terra gpt-5.6-terra	OpenAI	closed	88.0%
#4	Claude Mythos 5 claude-mythos-5	Anthropic	closed	88%
#5	GPT-5.6 Luna gpt-5.6-luna	OpenAI	closed	84.7%
#6	Claude Fable 5 claude-fable-5	Anthropic	closed	84.6%
#7	Claude Opus 4.8 claude-opus-4-8	Anthropic	closed	84.6%
#8	GPT-5.5 gpt-5.5	OpenAI	closed	84.3%
#9	Grok 4.5 grok-4.5	xAI	closed	83.3%
#10	GLM-5.2 glm-5-2	Z.AI	open	81%
#11	Claude Sonnet 5 claude-sonnet-5	Anthropic	closed	80.5%
#12	Muse Spark 1.1 muse-spark-1-1	Meta	closed	80%
#13	GPT-5.4 gpt-5.4	OpenAI	closed	78.3%
#14	Gemini 3.5 Flash gemini-3.5-flash	Google	closed	76.2%
#15	Qwen3.7-Max qwen3.7-max	Alibaba Cloud	—	74.5%
#16	Gemini 3 Pro gemini-3-pro	Google	closed	73.9%
#17	Gemini 3.1 Pro Preview gemini-3.1-pro-preview	Google	closed	73.8%
#18	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	70.3%
#19	Claude Opus 4.7 claude-opus-4-7	Anthropic	closed	69.4%
#20	MiMo-V2.5-Pro mimo-v2-5-pro	Xiaomi	closed	68.4%
#21	Kimi K2.7 Code kimi-k2.7-code	Moonshot AI	open	67.4%
#22	MiniMax M3 MiniMax-M3	MiniMax	—	66%
#23	Kimi K2.6 kimi-k2.6	Moonshot AI	open	65.9%
#24	Qwen3.6 Max qwen3.6-max	Alibaba Cloud	closed	65.4%
#25	DeepSeek V4 Pro deepseek-v4-pro	DeepSeek	open	64.0%
#26	MiMo-V2.5 mimo-v2.5	Xiaomi	open	63.7%
#27	DeepSeek V4 Flash deepseek-v4-flash	DeepSeek	open	61.8%
#28	Qwen3.6 27B qwen3.6-27b	Alibaba Cloud	open	60.7%
#29	GPT-5.4 mini gpt-5.4-mini	OpenAI	closed	59.2%
#30	GLM-5.1 glm-5-1	Z.AI	open	58.7%
#31	Claude Sonnet 4.5 claude-sonnet-4-5	Anthropic	closed	55.8%
#32	Hy3 hy3	Tencent	open	54.4%
#33	Nemotron 3 Ultra nemotron-3-ultra-550b	NVIDIA	open	53.9%
#34	Qwen3.5 397B qwen3.5-397b-a17b	Alibaba Cloud	open	51.3%
#35	Mistral Medium 3.5 mistral-medium-3-5	Mistral AI	open	50.6%
#36	GLM-4.6 glm-4.6	Z.AI	open	49.4%
#37	Qwen3.5 122B qwen3.5-122b-a10b	Alibaba Cloud	open	47.6%
#38	DeepSeek V3.2 deepseek-v3.2	DeepSeek	open	46.8%
#39	GLM-4.7 glm-4.7	Z.AI	open	45.3%
#40	DeepSeek V3.1 Terminus deepseek-v3.1-terminus	DeepSeek	open	44.9%
#41	Grok 4.3 grok-4.3	xAI	closed	39.7%
#42	Gemma 4 26B gemma-4-26b-a4b	Google	open	39.0%
#43	Nemotron 3 Super nemotron-3-super-120b	NVIDIA	open	38.6%
#44	Claude Sonnet 4 claude-sonnet-4	Anthropic	closed	36.3%
#45	GPT-5 gpt-5	OpenAI	closed	35.2%
#46	Nova 2 Pro nova-2-pro-preview	Amazon	closed	29.6%
#47	Gemini 2.5 Pro gemini-2.5-pro	Google	closed	28.5%
#48	Command A+ command-a-plus	Cohere	open	22.8%
#49	Mistral Small 4 mistral-small-2603	Mistral AI	open	21.0%
#50	Mistral Large 3 mistral-large-2512	Mistral AI	open	12.0%

The top of this snapshot is led by GPT-5.6 Sol at 91.9%; third place is 3.9 points behind. The top-10 spread is 10.9 points.

Compare top models

GPT-5.6 Sol vs Kimi K3 Kimi K3 vs GPT-5.6 Terra GPT-5.6 Terra vs Claude Mythos 5 Claude Mythos 5 vs GPT-5.6 Luna

About Terminal-Bench

Definition and scoring

Organisation: Terminal-Bench team
Category: Agents
Version: 2.1
Direction: higher is better
Ranking use: Weighted 1.6%
Contamination risk: Unknown

A benchmark for agents performing practical tasks in reproducible terminal 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 Terminal-Bench measure?

A benchmark for agents performing practical tasks in reproducible terminal environments.

Which model scores highest on Terminal-Bench?

GPT-5.6 Sol by OpenAI currently leads with 91.9%.

How many models are evaluated on Terminal-Bench?

50 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. GPT-5.6 Sol is overall #4.

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 · 24 results · Weighted 1.6%

OSWorld

Rolling · 11 results · Weighted 1.6%

WebArena

Rolling · 2 results · Not Weighted

Agents · Benchmark profile

Terminal-Bench

A benchmark for agents performing practical tasks in reproducible terminal environments.

Ranking weighted2.1Rolling50 models

Data verified 15 July 2026 · Methodology 1.2.0

Benchmark score on Terminal-Bench

Leaderboard · 50 models

GPT-5.6 Sol leads at 91.9%, followed by Kimi K3 (88.3%) and GPT-5.6 Terra (88.0%).

Ranking · 1.6%Agents2.1

#1Closed

GPT-5.6 Sol

OpenAI

gpt-5.6-sol

91.9%

Overall #4 · Context 1.05M tokens

#2Closed

Kimi K3

Moonshot AI

kimi-k3

88.3%

Overall #3 · Context 1M tokens

#3Closed

GPT-5.6 Terra

OpenAI

gpt-5.6-terra

88.0%

Overall #9 · Context 1.05M 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

OpenAIMoonshot AIAnthropicxAIZ.AIMetaGoogleAlibaba CloudXiaomi

Leaderboard bars

20 ranked models · higher is better

View accessible chart data

Accessible data for Leaderboard bars
Model	Rank	Provider	Score
GPT-5.6 Sol	#1	OpenAI	91.9%
Kimi K3	#2	Moonshot AI	88.3%
GPT-5.6 Terra	#3	OpenAI	88.0%
Claude Mythos 5	#4	Anthropic	88%
GPT-5.6 Luna	#5	OpenAI	84.7%
Claude Fable 5	#6	Anthropic	84.6%
Claude Opus 4.8	#7	Anthropic	84.6%
GPT-5.5	#8	OpenAI	84.3%
Grok 4.5	#9	xAI	83.3%
GLM-5.2	#10	Z.AI	81%
Claude Sonnet 5	#11	Anthropic	80.5%
Muse Spark 1.1	#12	Meta	80%
GPT-5.4	#13	OpenAI	78.3%
Gemini 3.5 Flash	#14	Google	76.2%
Qwen3.7-Max	#15	Alibaba Cloud	74.5%
Gemini 3 Pro	#16	Google	73.9%
Gemini 3.1 Pro Preview	#17	Google	73.8%
Qwen3.7-Plus	#18	Alibaba Cloud	70.3%
Claude Opus 4.7	#19	Anthropic	69.4%
MiMo-V2.5-Pro	#20	Xiaomi	68.4%

Score curve

Shows how quickly scores separate down the field

View accessible chart data

Accessible data for Score curve
Model	Rank	Provider	Score
GPT-5.6 Sol	#1	OpenAI	91.9%
Kimi K3	#2	Moonshot AI	88.3%
GPT-5.6 Terra	#3	OpenAI	88.0%
Claude Mythos 5	#4	Anthropic	88%
GPT-5.6 Luna	#5	OpenAI	84.7%
Claude Fable 5	#6	Anthropic	84.6%
Claude Opus 4.8	#7	Anthropic	84.6%
GPT-5.5	#8	OpenAI	84.3%
Grok 4.5	#9	xAI	83.3%
GLM-5.2	#10	Z.AI	81%
Claude Sonnet 5	#11	Anthropic	80.5%
Muse Spark 1.1	#12	Meta	80%
GPT-5.4	#13	OpenAI	78.3%
Gemini 3.5 Flash	#14	Google	76.2%
Qwen3.7-Max	#15	Alibaba Cloud	74.5%
Gemini 3 Pro	#16	Google	73.9%
Gemini 3.1 Pro Preview	#17	Google	73.8%
Qwen3.7-Plus	#18	Alibaba Cloud	70.3%
Claude Opus 4.7	#19	Anthropic	69.4%
MiMo-V2.5-Pro	#20	Xiaomi	68.4%

Provider profile

Average within the visible cohort · descriptive only

View accessible chart data

Accessible data for Provider profile
Provider	Models	Average score
Moonshot AI	1	88.3%
OpenAI	5	85.4%
xAI	1	83.3%
Anthropic	5	81.4%
Z.AI	1	81%
Meta	1	80%
Google	3	74.6%
Alibaba Cloud	2	72.4%
Xiaomi	1	68.4%

Benchmark score table (50 models)

One best score per model · higher is better

Terminal-Bench leaderboard ranked by score
Rank	Model	Provider	License	Score
#1	GPT-5.6 Sol gpt-5.6-sol	OpenAI	closed	91.9%
#2	Kimi K3 kimi-k3	Moonshot AI	closed	88.3%
#3	GPT-5.6 Terra gpt-5.6-terra	OpenAI	closed	88.0%
#4	Claude Mythos 5 claude-mythos-5	Anthropic	closed	88%
#5	GPT-5.6 Luna gpt-5.6-luna	OpenAI	closed	84.7%
#6	Claude Fable 5 claude-fable-5	Anthropic	closed	84.6%
#7	Claude Opus 4.8 claude-opus-4-8	Anthropic	closed	84.6%
#8	GPT-5.5 gpt-5.5	OpenAI	closed	84.3%
#9	Grok 4.5 grok-4.5	xAI	closed	83.3%
#10	GLM-5.2 glm-5-2	Z.AI	open	81%
#11	Claude Sonnet 5 claude-sonnet-5	Anthropic	closed	80.5%
#12	Muse Spark 1.1 muse-spark-1-1	Meta	closed	80%
#13	GPT-5.4 gpt-5.4	OpenAI	closed	78.3%
#14	Gemini 3.5 Flash gemini-3.5-flash	Google	closed	76.2%
#15	Qwen3.7-Max qwen3.7-max	Alibaba Cloud	—	74.5%
#16	Gemini 3 Pro gemini-3-pro	Google	closed	73.9%
#17	Gemini 3.1 Pro Preview gemini-3.1-pro-preview	Google	closed	73.8%
#18	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	70.3%
#19	Claude Opus 4.7 claude-opus-4-7	Anthropic	closed	69.4%
#20	MiMo-V2.5-Pro mimo-v2-5-pro	Xiaomi	closed	68.4%
#21	Kimi K2.7 Code kimi-k2.7-code	Moonshot AI	open	67.4%
#22	MiniMax M3 MiniMax-M3	MiniMax	—	66%
#23	Kimi K2.6 kimi-k2.6	Moonshot AI	open	65.9%
#24	Qwen3.6 Max qwen3.6-max	Alibaba Cloud	closed	65.4%
#25	DeepSeek V4 Pro deepseek-v4-pro	DeepSeek	open	64.0%
#26	MiMo-V2.5 mimo-v2.5	Xiaomi	open	63.7%
#27	DeepSeek V4 Flash deepseek-v4-flash	DeepSeek	open	61.8%
#28	Qwen3.6 27B qwen3.6-27b	Alibaba Cloud	open	60.7%
#29	GPT-5.4 mini gpt-5.4-mini	OpenAI	closed	59.2%
#30	GLM-5.1 glm-5-1	Z.AI	open	58.7%
#31	Claude Sonnet 4.5 claude-sonnet-4-5	Anthropic	closed	55.8%
#32	Hy3 hy3	Tencent	open	54.4%
#33	Nemotron 3 Ultra nemotron-3-ultra-550b	NVIDIA	open	53.9%
#34	Qwen3.5 397B qwen3.5-397b-a17b	Alibaba Cloud	open	51.3%
#35	Mistral Medium 3.5 mistral-medium-3-5	Mistral AI	open	50.6%
#36	GLM-4.6 glm-4.6	Z.AI	open	49.4%
#37	Qwen3.5 122B qwen3.5-122b-a10b	Alibaba Cloud	open	47.6%
#38	DeepSeek V3.2 deepseek-v3.2	DeepSeek	open	46.8%
#39	GLM-4.7 glm-4.7	Z.AI	open	45.3%
#40	DeepSeek V3.1 Terminus deepseek-v3.1-terminus	DeepSeek	open	44.9%
#41	Grok 4.3 grok-4.3	xAI	closed	39.7%
#42	Gemma 4 26B gemma-4-26b-a4b	Google	open	39.0%
#43	Nemotron 3 Super nemotron-3-super-120b	NVIDIA	open	38.6%
#44	Claude Sonnet 4 claude-sonnet-4	Anthropic	closed	36.3%
#45	GPT-5 gpt-5	OpenAI	closed	35.2%
#46	Nova 2 Pro nova-2-pro-preview	Amazon	closed	29.6%
#47	Gemini 2.5 Pro gemini-2.5-pro	Google	closed	28.5%
#48	Command A+ command-a-plus	Cohere	open	22.8%
#49	Mistral Small 4 mistral-small-2603	Mistral AI	open	21.0%
#50	Mistral Large 3 mistral-large-2512	Mistral AI	open	12.0%

The top of this snapshot is led by GPT-5.6 Sol at 91.9%; third place is 3.9 points behind. The top-10 spread is 10.9 points.

Compare top models

GPT-5.6 Sol vs Kimi K3 Kimi K3 vs GPT-5.6 Terra GPT-5.6 Terra vs Claude Mythos 5 Claude Mythos 5 vs GPT-5.6 Luna

About Terminal-Bench

Definition and scoring

Organisation: Terminal-Bench team
Category: Agents
Version: 2.1
Direction: higher is better
Ranking use: Weighted 1.6%
Contamination risk: Unknown

Open benchmark source ↗

FAQ

Quick answers

What does Terminal-Bench measure?

A benchmark for agents performing practical tasks in reproducible terminal environments.

Which model scores highest on Terminal-Bench?

GPT-5.6 Sol by OpenAI currently leads with 91.9%.

How many models are evaluated on Terminal-Bench?

50 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. GPT-5.6 Sol is overall #4.

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 · 24 results · Weighted 1.6%

OSWorld

Rolling · 11 results · Weighted 1.6%

WebArena

Rolling · 2 results · Not Weighted

Leaderboard · 50 models

GPT-5.6 Sol

Kimi K3

GPT-5.6 Terra

See the field, not just the table.

Benchmark score table (50 models)

Compare top models

Definition and scoring

Quick answers

What does Terminal-Bench measure?

Which model scores highest on Terminal-Bench?

How many models are evaluated on Terminal-Bench?

Does this affect overall Lumina rank?

More Agents benchmarks

Leaderboard · 50 models

GPT-5.6 Sol

Kimi K3

GPT-5.6 Terra

See the field, not just the table.

Benchmark score table (50 models)

Compare top models

Definition and scoring

Quick answers

What does Terminal-Bench measure?

Which model scores highest on Terminal-Bench?

How many models are evaluated on Terminal-Bench?

Does this affect overall Lumina rank?

More Agents benchmarks