Compare runs
Pick two or more runs to compare. Qwen3.8-27B-Fast on M4 Max 128 GB is the first; add another below.
No other run of Qwen3.8-27B-Fast or on M4 Max 128 GB to add yet.
| Run | Qwen3.8-27B-FastM4 Max 128 GB · llama.cppTake out Qwen3.8-27B-Fast |
|---|---|
| Score | 2,256Silver |
| Tasks passed | 49 of 50 |
| Points | 79 of 82 |
| Speed | 23.4 tok/s |
| Tokens | 100,191 |
| Summarizing | 5/5 |
| One-sentence summary | Passed25.3 tok/s |
| Meeting notes in three bullets | Passed25.1 tok/s |
| Review sentiment and complaint | Passed24.1 tok/s |
| Faithful summary of a study | Passed23 tok/s |
| Action items from an email thread | Passed25.1 tok/s |
| Extracting | 5/5 |
| Fields from an email as JSON | Passed24.5 tok/s |
| Every email address, once | Passed24.8 tok/s |
| A list as CSV | Passed24.9 tok/s |
| The real deadline | Passed24.8 tok/s |
| Warnings and errors from a log | Passed25.2 tok/s |
| Following instructions | 5/5 |
| Exactly one word | Passed25 tok/s |
| Three lowercase lines | Passed21.6 tok/s |
| Exact JSON shape | Passed24.9 tok/s |
| Sort a list | Passed24.6 tok/s |
| A name with three rules | Passed21.9 tok/s |
| Reasoning | 5/5 |
| Unit price | Passed20.1 tok/s |
| Day and date | Passed24.9 tok/s |
| Who owns each pet | Passed25.1 tok/s |
| Arrival time | Passed25 tok/s |
| Filling a tank | Passed24.7 tok/s |
| Reading code | 4/4 |
| What does it print | Passed20.9 tok/s |
| Trace a loop | Passed24 tok/s |
| Find the bug | Passed24.6 tok/s |
| Run a query in your head | Passed25.1 tok/s |
| Using tools | 5/5 |
| Call one tool | Passed24.4 tok/s |
| Don't call a tool when none is needed | Passed25.6 tok/s |
| Two calls for two answers | Passed25.4 tok/s |
| Chain two tools | Passed24.8 tok/s |
| Handle a failed call honestly | Passed24.3 tok/s |
| Building pages | 5/5 |
| A small website | Passed22.3 tok/s |
| A pricing table with a toggle | Passed24.9 tok/s |
| An accessible sign-up form | Passed25.3 tok/s |
| A to-do app that remembers | Passed25.3 tok/s |
| A bar chart in SVG | Passed24.4 tok/s |
| Building games | 3/4 |
| Build 2048 | Passed24.8 tok/s |
| Build Snake | Failed24.3 tok/s |
| Box-pushing puzzle | Passed23.1 tok/s |
| Minesweeper | Passed20 tok/s |
| Building apps | 4/4 |
| Step sequencer | Passed23.3 tok/s |
| Grow a branching plant | Passed20.8 tok/s |
| Falling sand | Passed23.8 tok/s |
| Focus timer | Passed23.9 tok/s |
| Writing code | 3/3 |
| Merge intervals | Passed25 tok/s |
| An LRU cache | Passed23.7 tok/s |
| Evaluate arithmetic | Passed24.6 tok/s |
| Fixing code | 2/2 |
| Fix the bookings module | Passed25.1 tok/s |
| Fix the CSV reader | Passed24.7 tok/s |
| Planning and data | 3/3 |
| A query over hidden data | Passed24.2 tok/s |
| Schedule five meetings | Passed24.5 tok/s |
| A replacement in ten calls or fewer | Passed24.1 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.