77.3%
▲ 58.9 pts since Feb 2025
top Terminal-Bench 2.0 accuracy, February 2026
20252026

What the chart shows

Terminal-Bench 2.0 tests AI agents on end-to-end tasks in a real terminal, from compiling code to training models and setting up servers, chaining steps without human help. The best score on its leaderboard rose from 18.4% in February 2025 to 77.3% in February 2026.

Progress came in steps: 32.6% from March to July 2025, 49.7% from August, 65.2% in November and 66.5% in December, before the rise to 77.3%. The chart shows the leading score each month, not one model over time.

Key numbers

MeasureValue
February 202518.4%
March 202532.6%
August 202549.7%
November 202565.2%
December 202566.5%
February 202677.3%

How to read it

  • Each point is the highest accuracy on the leaderboard in that month; the leading agent and model change over time.
  • The report text gives the February 2025 starting point as 20%; the data shows 18.4%, which we use.
Source: Terminal-Bench 2.0 Leaderboard, 2026, via the AI Index Report 2026 (Stanford HAI), figure 2.5.2: Terminal-Bench 2.0: accuracy. Chart and text by AI Adoption.
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart