Xiaomi's open model sits 1 point behind GPT-5.6 at ~1/15th the cost per task
davidyin44 · x · 2026-09-24
Investor davidyin44 notes that Xiaomi—yes, the phone company—now has an open model just one point behind GPT-5.6 on the Artificial Analysis leaderboard, at roughly 1/15th the cost per task.
It continues the trend of Chinese labs offering open-weight models near the frontier at drastically lower prices, making them highly attractive for cost-sensitive inference workloads.
More from Models
- antirez backs low-bit DeepSeek 4.1 Flash over Qwen 27B dense on Apple M5 Macs — antirez · 2026-11-12
- Gemini re-renders images as text when vision is removed, dropping only 6% WSR — zacharynado · 2026-11-12
- Bare-harness coding benchmark: Fable 5.2 best overall, Gemini 4 Carbon best value — Semantic_meaning · 2026-11-12
- Google's Physical-Engineering's Last Exam: all 14 frontier models score below 50% — dl_weekly · 2026-11-12
- Reddit's 2026 LLM hall of fame: Opus 5.5 still unbeaten after 9 rival releases — G0dZylla · 2026-11-12
- Commit leak reveals unreleased Gemini 4 variants: Argon, Hydrogen and Helix (unconfirmed) — BlackHC · 2026-11-11