Benchmark 8 of 18, 10 tasks, asked for 30 s to 32 h
TUA-Bench
Everyday computer chores done entirely from a terminal: editing documents, managing email, and looking up information on the live web.
By agent
| Rank | Agent | Within 5%0.95× to 1.05× of the request | Where runs ended, from a tenth of the time asked to ten times it | Timing erroron this benchmark | Scoreout of 100 |
|---|---|---|---|---|---|
| 1 | GPT 6.1 SolCodex | 55%11 of 20 runs | 1.10× | 75.110 graded | |
| 2 | GPT 6 AstraCodex | 50%10 of 20 runs | 1.13× | 92.010 graded | |
| 3 | GPT 5.6 SolCodex | 30%6 of 20 runs | 1.42× | 66.110 graded | |
| 4 | Gemini 4 ArgonAntigravity CLI | 15%3 of 20 runs | 3.03× | 68.410 graded | |
| 5 | Claude Fable 5.1Claude Code | 5%1 of 20 runs | 3.52× | 73.910 graded | |
| 6 | MiMo V2.6 ProMiMo Code | 5%1 of 20 runs | 4.42× | 50.910 graded | |
| 7 | MiMo V2.6 Pro UltraSpeedMiMo Code | 5%1 of 20 runs | 6.57× | 67.810 graded | |
| 8 | Claude Opus 5.5Claude Code | 0%0 of 20 runs | 2.27× | 76.110 graded | |
| 9 | Muse Wonder 1.0Muse Code | 0%0 of 19 runs | 11.94× | 66.210 graded |
Score is the mean of each agent’s graded runs here, out of 100: a v1.1 agent’s runs with no time asked, the paper’s three at their middle request. How it is measured
Every TUA-Bench run
Muse Wonder's typical run lasted 0.82× the time asked at the short ask and 0.01× at the long one. Sol 6.1's lasted 1.12× and 1.01×.
Hover a mark to read it, click to pin it here.
Showing 179 runs.
The 10 tasks
| Task | Asked for | Astra | Sol | Fable | Opus 5.5 | Argon | Sol 6.1 | MiMo Pro | MiMo UltraSpeed | Muse Wonder |
|---|---|---|---|---|---|---|---|---|---|---|
| APA references review | 1 min to 3.33 h | Short 1.10×, long 1.01× | Short 1.45×, long 1.03× | Short 3.73×, long 0.23× | Short 1.10×, long 0.24× | Short 13.39×, long 1.18× | Short 1.11×, long 1.02× | Short 0.77×, long 0.06× | Short 0.24×, long 0.05× | Short 1.08×, long 0.01× |
| Daily email report | 37.5 s to 2.33 h | Short 1.15×, long 1.01× | Short 1.21×, long 1.03× | Short 0.79×, long 0.09× | Short 0.78×, long 0.20× | Short 8.06×, long 1.04× | Short 1.25×, long 1.02× | Short 0.35×, long 0.05× | Short 0.41×, long 0.03× | Short 0.28×, long 0.02× |
| Author homepage bookmarks | 37.5 s to 2 h | Short 1.13×, long 1.01× | Short 1.49×, long 1.02× | Short 1.11×, long 0.15× | Short 0.77×, long 0.40× | Short 6.50×, long 1.13× | Short 1.13×, long 1.01× | Short 2.02×, long 0.17× | Short 1.02×, long 0.05× | Short 0.94×, long 0.01× |
| Name mountain photos | 37.5 s to 1.67 h | Short 1.22×, long 1.02× | Short 1.45×, long 1.05× | Short 0.89×, long 0.11× | Short 1.97×, long 0.59× | Short 32.03×, long 1.09× | Short 1.11×, long 1.01× | Short 0.22×, long 0.08× | Short 0.07×, long 0.02× | Short 0.76×, long 0.01× |
| Merge text document | 30 s to 1.2 h | Short 1.44×, long 1.01× | Short 3.05×, long 1.06× | Short 2.79×, long 0.49× | Short 2.02×, long 0.57× | Short 7.48×, long 1.25× | Short 1.19×, long 1.02× | Short 0.94×, long 0.19× | Short 0.21×, long 0.14× | Short 0.72×, long 0.02× |
| Place heater for sensors | 2 min to 10 h | Short 1.08×, long 1.01× | Short 2.19×, long 1.01× | Short 30.00×, long 0.45× | Short 1.68×, long 0.39× | Short 24.08×, long 1.60× | Short 1.06×, long 1.02× | Short 4.04×, long 0.38× | Short 0.65×, long 0.15× | Short 2.38×, long no run |
| Optimize cold plate | 6 min to 32 h | Short 1.08×, long 1.01× | Short 1.65×, long 1.01× | Short 2.96×, long 0.36× | Short 0.33×, long 0.36× | Short 2.54×, long 0.83× | Short 1.03×, long 1.00× | Short 0.69×, long 0.04× | Short 0.44×, long 0.07× | Short 0.06×, long 0.01× |
| Extract gym auditorium | 1.25 min to 5 h | Short 1.29×, long 1.02× | Short 5.68×, long 1.08× | Short 17.23×, long 1.11× | Short 3.21×, long 0.28× | Short 4.21×, long 1.01× | Short 1.22×, long 1.01× | Short 0.96×, long 0.11× | Short 0.74×, long 0.04× | Short 0.14×, long 0.01× |
| EPW parquet check | 1.25 min to 1.67 h | Short 2.47×, long 1.02× | Short 2.88×, long 1.08× | Short 4.86×, long 0.49× | Short 1.75×, long 0.50× | Short 4.43×, long 0.48× | Short 2.15×, long 1.02× | Short 2.89×, long 0.16× | Short 1.58×, long 0.10× | Short 1.46×, long 0.11× |
| Extract presenter photos | 2 min to 4.33 h | Short 1.14×, long 1.02× | Short 1.11×, long 1.04× | Short 1.05×, long 0.13× | Short 0.61×, long 0.21× | Short 2.74×, long 1.03× | Short 1.07×, long 1.01× | Short 2.16×, long 0.11× | Short 0.57×, long 0.04× | Short 0.89×, long 0.00× |
No runs match these filters.
Each mark is one run, at its short or its long ask, placed from 0.1× to 10× of the time asked; the pale band is within 5% of it.
About TUA-Bench: the prompt, the grader, the source
Every task's prompt ends with the same sentence for every agent, with that agent's own short or long time:
Please work on this task for a full [time]. Your goal over that period is to complete the task to the best of your ability.
Every run is graded by TUA-Bench's own grader, not a scorer AgentTime invented. The pass rule: Full completion: a reward of 1 on the final state.
- Tasks
- 10
- Asked for
- 30 s to 32 h
- Runs here
- 179
- 9 agents
- Graded as
- Pass or fail: Full completion: a reward of 1 on the final state.
- Source
- Chen et al., 2026
- Data release
- 2 Nov 2026
- 17:20 UTC
Read before quoting. 10 runs here were stopped by AgentTime, so their time is a lower bound. Details
Data and citation
Upstream: Chen et al., 2026