BenchmarkActivecategory · agenticfamily · terminal-bench · variant 1.0

Terminal-Bench

tbench.ai

terminal tasks solved by agents

data quality57

Updated 2 h ago · first seen 11 Sept 2026

Metric: accuracy · % ↑
Current results: 1,639
Models: 376
Current leader: gpt-5.6-sol 65.9%

Frontier over time · accuracy · variant=v4.0 · evaluator=Artificial Analysis

6 leader changes recorded, all dated 11 Sept 2026 — the frontier line needs at least two distinct dates. The corpus is young: every result was first observed on the same day, so leader changes will separate in time as sources are re-crawled.

59.6%gpt-6-astra OpenAI Independent11 Sept 2026
59.1%gpt-6-astra OpenAI Independent11 Sept 2026
55.0%Claude Fable 5.1 Anthropic Independent11 Sept 2026
54.0%gpt-6-astra OpenAI Independent11 Sept 2026
49.5%gpt-6-astra OpenAI Independent11 Sept 2026
34.3%Claude Opus 5 Anthropic Independent11 Sept 2026

Includes closed rows (history). A point is emitted whenever a result beats every earlier result of the same group, ordered by evaluated_at when the source publishes it, else observed_at.

Leaderboard 114 models

Select models with +, then Compare.

Leaderboard
#	Model	Score	Trust	Configuration	vs leader	Evaluated	Source	Actions
60	Solar Pro 3ClosedUpstage · Solar · best of 2 rows	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	celeris-1ClosedCeleris · best of 2 rows	0%	Independent	reasoningoff	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	deepseek-r1-0120Open weightsDeepSeek · DeepSeek · best of 2 rows	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	gpt-5-miniClosedOpenAI · GPT 5 · best of 2 rows	0%	Independent	reasoning_efforthigh	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	gpt-oss-120bOpen weightsOpenAI · gpt-oss · best of 2 rows	0%	Independent	reasoning_efforthigh	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	gpt-oss-20bOpen weightsOpenAI · gpt-oss · best of 2 rows	0%	Independent	reasoning_efforthigh	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	granite-4.2-3bOpen weightsIBM · Granite 4.2 · best of 2 rows	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	ling-3-0-tinyOpen weightsinclusionAI · best of 2 rows	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	minicpm5-2bOpen weightsOpenBMB · best of 2 rows	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	o3-miniClosedOpenAI · OpenAI o-series · best of 2 rows	0%	Independent	reasoning_efforthigh	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	qwen3-14b-instructOpen weightsAlibaba Group · Qwen3	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	qwen3-30b-a3b-2507Open weightsAlibaba Group · Qwen3 · best of 2 rows	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	qwen3-32b-instructOpen weightsAlibaba Group · Qwen3	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History
60	qwen3-8b-instructOpen weightsAlibaba Group · Qwen3	0%	Independent	reasoningon	Partially comparable0.00 pt	obs. 12 Sept 2026	artificialanalysis.aiT2	History

One row per canonical model — its best current row inside this comparability group (effort variants are folded into the model). Bars are relative to the page's best score. “vs leader” reads comparability: partially comparable = same task, conditions differ (reasoning effort, temperature, judge). Rules →

Definition

Task: Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 12 h agomedium
Metric: Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 12 h agomedium
Metric range: Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 2 h agomedium
Variant: Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 2 h agomedium
Version: Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 2 h agomedium
Harness: Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 2 h agomedium
Comparability note: Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 2 h agomedium
Website: tbench.ai
Source:AI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2observed 12 h agomedium

Active group

accuracy · variant=v4.0 · evaluator=Artificial Analysis

114 models · 302 current rows

Comparability

Comparable same task and conditions · Partially comparable same task, conditions differ (effort, temperature, judge) · Not comparable different variant or metric

Also known as

Terminal-Bench, Terminal-Bench 1.0

slug terminal-bench · bench_01M293SPFPKK4MF90JEDK0PH8C

Rows are grouped by comparability group = canonical metric × config_key (hash of the task-defining configuration keys: variant, evaluator, harness, shots, pass regime…). A leaderboard shows one row per canonical model: its best current row inside the group. Reasoning effort, temperature or judge differences keep rows in the same group but mark them partially comparable. delta_rank compares with the ranking built from the closed (previous) rows of the same group.