Hermes · claude-haiku-4-5-20251001 · as published
anthropic
as publishedIts score and figures
Every figure the leaderboard's table prints for this setup.
| rank in group | 1–12 |
|---|---|
| Score | 69.8 pts62.8–76.7 pts |
| Pass rate | 79.7%71.0–88.4%79.7% · 69 of 69 graded |
| Efficiency | 39.936.1–42.0 |
| cost | 0.45× |
| tokens | 1.1× |
| time | 1.6× |
| Coverage | 1/6 |
| Runs | 69 |
59.8 + 6.5 + 1.7 + 1.8 = 69.8 (outcome + cost + tokens + time)
- pass rate on each benchmark's own grader
- cost
- tokens
- time
- not measured
- First run
- 2026-10-03
- Last run
- 2026-10-03
- did not finish (scored as a fail)
- 0
- Tasks with no grade
- 0
On each benchmark
Each benchmark it ran: its status, pass rate, spend and spend per pass. The studies behind them are its evidence, below.
IFEval
69.8 pts
IFEval: 79.7% · 69 of 69 graded · scored
- cost
- 0.45× the typical spend per passing run25.2% of this setup's cost here went to runs that did not pass or were not graded; efficiency does not price it$0.01 per pass($0.0050–$0.01)29% went to the costliest 10% of attempts (7 of 69)
- tokens
- 1.1× the typical spend per passing run26.2% of this setup's tokens here went to runs that did not pass or were not graded; efficiency does not price it21.6k per pass(18k–25.8k)18% went to the costliest 10% of attempts (7 of 69)cache split not recorded on 69 attempts
- time
- 1.6× the typical spend per passing run21.4% of this setup's time here went to runs that did not pass or were not graded; efficiency does not price it40.1s per pass(35.4s–45.8s)14% went to the costliest 10% of attempts (7 of 69)
Efficiency and spend per pass can order setups differently: efficiency compares the typical (median) passing run, task by task, while spend per pass averages every attempt, so a few very expensive runs, passing or not, count in full; and spend per pass also counts the attempts that failed or did not finish, and divides by the pass rate. A setup is called the most frugal only where both put it first and both separate it from every other setup.
Evidence
Every published study its runs come from.
- ifeval-13-harnesses-2026-10 (69 runs)