Unsloth quants run 1T-param Mistral Large 4 on Mac Studio: 1-bit keeps 91% accuracy at 223GB
danielhanchen · x · 2026-10-29
Unsloth's Daniel Han released quants for Mistral Large 4 (Le Chonk) with multi-domain testing:
- 1-bit retains 91% of top-1 accuracy across coding, agentic tasks, math and more, while being 80% smaller (223GB vs 1.1TB FP8)
- The 1T-parameter model runs locally on a Mac Studio via 1-bit GGUF; claimed as the strongest Western open model to date
- All quants work in Unsloth Desktop with day-zero support from Mistral; KLD benchmark curves cover chat, math, reasoning, coding, tool calling, agentic coding, and multilingual
Related event: Unsloth Releases Le Chonk Quantized Models for Local Use(2 posts)→
More from Infra
- Microsoft, Alphabet, Amazon to spend ~$600B on AI infrastructure in 2026 — PeterDiamandis · 2026-10-29
- Sam Altman wants OpenAI robots to build data centers for AI's own expansion — VraserX · 2026-10-29
- Dual 3090s running Qwen3.8-27B INT4 is significantly faster than FP8 — Travnewmatic · 2026-10-29
- Junction: Two-Week Chip Fab Retreat in Bangalore to Train India's Semiconductor Talent — itsarnavb · 2026-10-29
- Only 13% of Planned UK Data Centre Capacity Under Construction, Report Finds — ersatzben · 2026-10-29
- SMCI-backed neocloud Argentum AI seeks $5B credit facility for chip purchases — Beth_Kindig · 2026-10-29