Observed arrival · 2026-11-06
Neyra tests the changes you make to local language models
Neyra is an open-source command-line toolkit for checking training data, tuning or editing local language models, and measuring the effects.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Local-model developers evaluating fine-tunes and edits
- Worth noticing
- Its public-tune test reports 98.5% on 200 banking questions, versus 90% with 40 prompt examples; the page notes the task is narrow.
Field notes
The benchmark display compares a base model, a 40-example prompt, a tuned model, and a quantized GGUF file, and includes a confidence interval plus cases the tune broke. A separate fact-edit example changes “the capital of France” to Rome, then reports exact reversion while failing a nearby-fact specificity check. The homepage says its current tests focus on smaller models, mostly Qwen, and that training presently runs on Apple Silicon.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue