Terminal-Bench Hard-AA

Terminal-Bench Hard (AA subset of Terminal-Bench 2.0): agentic terminal tasks; pass rate.

124 models · 194 results · data as of 2026-11-06 · artificialanalysis.ai ↗

More from this benchmark

3 published measures

Score vs any published measure3 measures: sub-scores, splits, counts

Highlight models

Click a point on the chart to highlight that exact effort; click it again to remove it.

Chart settings

124 of 124 selected

79 more models under All models in settings.

darker is better within each column · click a column to sort

194 rows

#	Model	Effort	Terminal-Bench Hard-AA
1	GPT-5.6 Sol	max	65.9%
2	Fable 5	max	62.9%
3	GPT-5.6 Sol	medium	62.9%
4	GPT-5.6 Terra	xhigh	62.9%
5	GPT-5.6 Sol	high	62.1%
6	GPT-5.6 Sol	xhigh	61.4%
7	GPT-5.5	xhigh	60.6%
8	GPT-5.6 Sol	low	60.6%
9	GPT-5.5	high	59.8%
10	Opus 4.8	max	58.3%
11	GPT-5.4	xhigh	57.6%
12	GPT-5.5	medium	57.6%
13	GPT-5.6 Terra	high	57.6%
14	GPT-5.6 Terra	max	57.6%
15	Opus 4.7	high	54.5%
16	Gemini 3.1 Pro		53.8%
17	GPT-5.3 Codex	xhigh	53.0%
18	GPT-5.4 Mini	xhigh	52.3%
19	GPT-5.5	low	52.3%
20	Opus 4.7	max	51.5%
21	GLM 5.2	max	50.8%
22	Qwen 3.7 Max	max	50.8%
23	GPT-5.5	none	49.2%
24	Opus 4.6	high	48.5%
25	Opus 4.5		47.0%
26	GPT-5.2	xhigh	47.0%
27	Qwen 3.7 Plus		47.0%
28	Sonnet 4.6	none	46.2%
29	DeepSeek V4 Pro 0424	max	46.2%
30	Gemini 3.5 Flash	minimal	46.2%