Speed and memory priorities require Pi 5 measurements. Set both to Ignore to compare all quality-tested models.
Set at least one priority to Low or higher to rank models.
Text input is included. Select every additional input type you need.
Only tested configurations are available. Longer contexts can use more memory.
Only includes models tested with reasoning switched off.
Input types come from the model publisher. This picker ranks text results; see the Vision and Audio tabs for those test scores.
Only models with measured scores in this language can match.
Preset voices are benchmarked. Voice cloning and naturalness ratings are not measured yet.
A limit requires an actual memory measurement on your Raspberry Pi 5.
Without device results, the picker can show a quality shortlist. Pi 5 speed, memory and fit remain unverified.
These tests cover specific capabilities. Tool calling does not measure a complete browsing agent; RULER does not measure every form of long context reasoning.
Enable JavaScript to use the picker, or compare the text results, speech recognition results and speech generation results directly.