Compare runs
Pick two or more runs to compare. OpenAI-gpt-oss-120b on an unnamed computer is the first; add another below.
| Run | OpenAI-gpt-oss-120bNo computer named · llama-swapTake out OpenAI-gpt-oss-120b |
|---|---|
| Score | 7,762Platinum |
| Tasks passed | 43 of 44 |
| Points | 63 of 64 |
| Speed | 78.8 tok/s |
| Tokens | 44,205 |
| Summarizing | 4/5 |
| One-sentence summary | Failed79.6 tok/s |
| Meeting notes in three bullets | Passed79.7 tok/s |
| Review sentiment and complaint | Passed79.6 tok/s |
| Faithful summary of a study | Passed79.8 tok/s |
| Action items from an email thread | Passed79.7 tok/s |
| Extracting | 5/5 |
| Fields from an email as JSON | Passed79.7 tok/s |
| Every email address, once | Passed79.6 tok/s |
| A list as CSV | Passed79.5 tok/s |
| The real deadline | Passed79.7 tok/s |
| Warnings and errors from a log | Passed79.4 tok/s |
| Following instructions | 5/5 |
| Exactly one word | Passed79.5 tok/s |
| Three lowercase lines | Passed80 tok/s |
| Exact JSON shape | Passed79.2 tok/s |
| Sort a list | Passed80 tok/s |
| A name with three rules | Passed79.8 tok/s |
| Reasoning | 5/5 |
| Unit price | Passed79.5 tok/s |
| Days of the week | Passed79.7 tok/s |
| Who owns the cat | Passed79.8 tok/s |
| Arrival time | Passed79.1 tok/s |
| Filling a tank | Passed79.5 tok/s |
| Reading code | 4/4 |
| What does it print | Passed79.8 tok/s |
| Trace a loop | Passed79.6 tok/s |
| Find the bug | Passed79.8 tok/s |
| Run a query in your head | Passed79.1 tok/s |
| Using tools | 5/5 |
| Call one tool | Passed79 tok/s |
| Don't call a tool when none is needed | Passed79.5 tok/s |
| Two calls for two answers | Passed78.8 tok/s |
| Chain two tools | Passed79.1 tok/s |
| Handle a failed call honestly | Passed79.4 tok/s |
| Building pages | 5/5 |
| A small website | Passed79.5 tok/s |
| A pricing table with a toggle | Passed79.2 tok/s |
| An accessible sign-up form | Passed79.3 tok/s |
| A to-do app that remembers | Passed79.2 tok/s |
| A bar chart in SVG | Passed79.2 tok/s |
| Building games | 2/2 |
| Build 2048 | Passed77.8 tok/s |
| Build Snake | Passed78.3 tok/s |
| Writing code | 3/3 |
| Merge intervals | Passed79.2 tok/s |
| An LRU cache | Passed79.2 tok/s |
| Evaluate arithmetic | Passed78.7 tok/s |
| Fixing code | 2/2 |
| Fix the bookings module | Passed78.9 tok/s |
| Fix the CSV reader | Passed78.9 tok/s |
| Planning and data | 3/3 |
| A query over hidden data | Passed79.2 tok/s |
| Schedule five meetings | Passed78.2 tok/s |
| A replacement in ten calls or fewer | Passed78.5 tok/s |
Scrolls sideways to show every run.
A check marks the best of each row. tok/s is tokens a second: how fast the model writes. How the score works.