Strata's IQ3_S quant rated on par with K4_XL, suggesting near-lossless Q2.5 compression
lxfater · x · 2026-10-07
Kaitchup, an account focused on evaluating post-quantization agent capabilities, found Strata's most popular IQ3S quant on par with K4XL — no loss, no input growth — while K4XL is widely seen as near-lossless vs the original. That implies much of today's model files can be compressed to roughly Q2.5 (IQ3S is a Q2.5-Q4 mix) with minimal loss; lxfater says he's already using it happily.
More from Models
- Rumor: OpenAI is sitting on two more batches of AI-generated math proofs — basedjensen · 2026-10-07
- Claude Opus 5.5 Builds a 600-Part Turbofan Engine From Public Data — and Simulates the Airflow — bookwormengr · 2026-10-07
- Mistral Large 4 posts strong STEMBench score across six research domains — GuillaumeLample · 2026-10-07
- Mistral releases 1T-parameter Large 4, nicknamed Le Chonk, to challenge top US and Chinese models — Ars Technica AI · 2026-10-07
- Google prototyping early voice agent codenamed "Concierge" alongside Gemini 4 prep — testingcatalog · 2026-10-07
- Mistral Large 4 claims #1 open-source model on Harvey's legal agent benchmark — MistralAI · 2026-10-07