AgenC Core: 76.2% on Terminal-Bench 2.1 for $34
A public run with DeepSeek V4.1 Flash. Every entry on the Harbor Hub leaderboard spent at least $134.
AgenC Core scored 76.2% on Terminal-Bench 2.1 with DeepSeek V4.1 Flash, and the whole run cost $34. On the Harbor Hub leaderboard, every entry spent at least $134 (leaderboard data from October 14, 2026).
How we ran it
We used the official Terminal-Bench 2.1 package: 89 tasks with 5 trials each, 445 trials in total, under the official timeouts and resources. When the agent looked up an answer online, we counted the trial as a failure. All 445 trials are public, with full trajectories.
This is a public run, not a leaderboard entry. It used AgenC Core's light mode, which is still experimental (agenc --light).
Why it costs so little
Light mode keeps the prompt lean and its opening stable, so DeepSeek serves most of it from cache.
The chart, the cost of each run and the trial data are on get.agenc.ag.
