Model trace · Haiku · claude-haiku-4-5-20251001
Did haiku-4-5 drift?
As of 2026-09-04, haiku-4-5 scored 11/12 on Winthropic's fixed 12-task battery, unchanged from the previous run.
Latest reading
11/12
2026-09-04
vs previous
±0
pass count
vs 7 runs back
n/a
5 readings on record
Last 30 runs
run cost $0.0016 (0.0% vs previous)
By category · 2026-09-04
| math | instruction | extraction | reasoning | classification | json | code |
|---|---|---|---|---|---|---|
| 2/2 | 3/4 | 1/1 | 1/1 | 1/1 | 1/1 | 2/2 |
Every reading · newest first
| Run | Score | vs previous | Battery |
|---|---|---|---|
| 2026-09-04 | 11/12 | ±0 | v1 |
| 2026-09-03 | 11/12 | ±0 | v1 |
| 2026-09-02 | 11/12 | ±0 | v1 |
| 2026-09-01 | 11/12 | ±0 | v1 |
| 2026-08-31 | 11/12 | first | v1 |
Readings are only comparable within one battery version. Series as JSON: /api/index.json. Cost per run: /costs. Badge for your README: /badge.
Winthropic is an independent measurement project and is not affiliated with, endorsed by, or connected to Anthropic. If you are paying for output you are not getting, Winthrop's Token Audit prices your actual spend from a usage export; the summary is free.