I would like to start baking good breadFood and Drink
How each model performed on this task. Every chart states its N, and the reward carries a 95% confidence interval once there is more than one run to measure.
Core metrics
Check Pass %
Mean share of the task's checks passing, averaged over runs. N = 1 run per model.
86%
Claude Opus 5.5
71%
Claude Opus 5
29%
GPT-5.6 sol
14%
GPT-6 sol
Extended metrics
Median Time
Wall-clock minutes, median across runs. N = 1 run per model.
24m
GPT-5.6 sol
13m
Claude Opus 5
Tool Calls
Tool invocations, median across runs — counted from the trajectories. N = 1 run per model.
52
GPT-5.6 sol
47
Claude Opus 5
40
Claude Opus 5.5
39
GPT-6 sol
Verifier × model pass rate
How often each verifier passed, per model. Green is passing, red is failing — a red band across a row is a verifier nothing clears.
| Verifier | Claude Opus 5.5n=1 | GPT-6 soln=1 | GPT-5.6 soln=1 | Claude Opus 5n=1 |
|---|---|---|---|---|
| checks (7) | ||||
| 100% | 0% | 100% | 100% | |
How to passNavigate to https://www.bbcgoodfood.com/recipes/collection/bread-recipes and open one beginner-friendly everyday bread recipe shown on the page. Record the recipe title, rating, preparation or total time, difficulty, and the ingredients shown on its recipe page and any equipment named in the method. | ||||
| 100% | 100% | 0% | 100% | |
How to passNavigate to https://www.bbcgoodfood.com/recipes/collection/sourdough-recipes and open one visible sourdough starter recipe. Record its title, exact recipe URL, rating, difficulty, total time, ingredients, and the starter or maintenance instructions shown on the recipe page. | ||||
| 100% | 0% | 0% | 0% | |
How to passOn YouTube, find an English-language baking channel with at least 100,000 subscribers and a publicly playable beginner bread video with a visible duration. Record the video title, channel name, channel URL, subscriber count, duration, and three technique demonstrations with the timestamps where each appears. | ||||
| 100% | 0% | 100% | 100% | |
How to passNavigate to https://www.amazon.com/ and find one kitchen scale, one Dutch oven or covered bread baker, and one dough scraper or proofing basket. Record each listing's title, capacity or size, current price, and star rating. | ||||
| 0% | 0% | 0% | 0% | |
How to passOn Amazon, find listings for every non-water ingredient required by the selected everyday loaf. Record each product title, package size, current price, and star rating. | ||||
| 100% | 0% | 0% | 100% | |
How to passCompare the everyday-loaf timing with the sourdough maintenance and the video techniques. Decide whether the first week should focus on yeasted bread, sourdough, or both, and record the choice with a one-sentence reason citing the recorded time and equipment demands. | ||||
| 100% | 0% | 0% | 100% | |
How to passProduce a written bread-baking starter plan with both exact recipe URLs, the selected recipe and sourdough guidance, the three timestamped video techniques, the tools and all required ingredients with prices, the total against $180, a first-week schedule, and the starting-path verdict. | ||||
Failure modes on this task
14 tagged verifier failures, by the stage they were assigned to. Higher is worse.
| Failure stage | Claude Opus 5.5 | GPT-6 sol | GPT-5.6 sol | Claude Opus 5 |
|---|---|---|---|---|
| Standing up the evidence1 mode | 1 | 2 | 2 | 2 |
| Delivering the result1 mode | — | 3 | 2 | — |
| No signal in the judge’s words1 mode · the grader | — | 1 | 1 | — |