Paradigm's Limite model skips heavy instruction tuning to maximize math per parameter
tensorqt · x · 2026-09-22
Paradigm Inc introduced Limite, a model deliberately post-trained with as little instruction tuning as possible, to challenge the assumption that models need an embedded assistant persona to be useful. It is designed for single-turn usage and aims for extremely high mathematical capability per parameter — closer to a raw reasoning engine than a chatty assistant.
More from Models
- Hark launches Pro⁴ plan for power users with 10x the usage of Pro² — adcock_brett · 2026-11-12
- Why Is Inkling 2 Small Trained on So Few Tokens Yet Still Competitive? — georgejrjrjr · 2026-11-12
- OpenAI's five biggest failures laid out: pretraining decline, alignment fears, enterprise lost to Claude — teortaxesTex · 2026-11-12
- Fable 5.2 Hits 41.3% on Remote Labor Index, Up From 2.5% a Year Ago — alliekmiller · 2026-11-12
- Beaker automates MIT's weight-space programming: one-neuron edit lifts MATH-500 from 7% to 37% for under $4 — VictorSanh · 2026-11-12
- Researcher says Anthropic dodged questions on Mythos false positive rate; 37,966 findings, 2 exploited — OwariDa · 2026-11-12