resyn

Compare runs

Pick two or more runs to compare. Llama-3.2-3B-Uncensored on an unnamed computer is the first; add another below.

Runs of the same model, or on the same computer, the newest first.

RunLlama-3.2-3B-UncensoredNo computer named · Ollama v262Take out
Score—No score
Tasks passed9 of 44
Points9 of 64
Speed67 tok/s
Tokens9,484
Summarizing2/5
One-sentence summary66.4 tok/s
Meeting notes in three bullets67.2 tok/s
Review sentiment and complaint66.4 tok/s
Faithful summary of a study67.4 tok/s
Action items from an email thread67 tok/s
Extracting3/5
Fields from an email as JSON67.3 tok/s
Every email address, once
A list as CSV67.2 tok/s
The real deadline66.8 tok/s
Warnings and errors from a log67 tok/s
Following instructions2/5
Exactly one word64.4 tok/s
Three lowercase lines66.8 tok/s
Exact JSON shape67 tok/s
Sort a list66.8 tok/s
A name with three rules66.5 tok/s
Reasoning0/5
Unit price
Days of the week
Who owns the cat
Arrival time
Filling a tank
Reading code0/4
What does it print
Trace a loop
Find the bug66.5 tok/s
Run a query in your head
Using tools0/5
Call one tool66.9 tok/s
Don't call a tool when none is needed
Two calls for two answers
Chain two tools
Handle a failed call honestly66.9 tok/s
Building pages2/5
A small website67.2 tok/s
A pricing table with a toggle67.1 tok/s
An accessible sign-up form67 tok/s
A to-do app that remembers67.1 tok/s
A bar chart in SVG
Building games0/2
Build 2048
Build Snake
Writing code0/3
Merge intervals
An LRU cache
Evaluate arithmetic
Fixing code0/2
Fix the bookings module
Fix the CSV reader
Planning and data0/3
A query over hidden data
Schedule five meetings
A replacement in ten calls or fewer

A check marks the best of each row. tok/s is tokens a second: how fast the model writes. How the score works.