AI-designed model defies human intuition: 40B total params, 2.5B active, 4-bit LM head
thursdai_pod · x · 2026-10-26
Recursive's Eureka Machine, an autonomous AI scientist, picked an architecture human researchers would likely reject for a local coding model: 40B total params with only 2.5B active, plus a 4-bit quantized LM head.
Co-founder Tim Shi explains on the ThursdAI podcast why the counterintuitive design works well for on-device coding workloads.
More from coding & agent
- Fewer, well-designed agent harnesses beat sprawling agent fleets, argues vixsheikh — annetgriffin · 2026-10-26
- ConformWeb: LLM-generated web apps pass only 11.7% of behavioral contract checks — algo_diver · 2026-10-26
- Agent memory explained in 12 minutes: working, procedural, semantic and episodic — _AustinCalvert_ · 2026-10-26
- Grok Bot v0.75.5 leak touts stored cards, email forwarding, 3x automation limits — mark_k · 2026-10-26
- Every time my agent breaks my UI, I add a rule to its skill — 10 design laws — _AustinCalvert_ · 2026-10-26
- Turn nouns into verbs: the 'Range' trick to unlock cross-domain AI output — petesena · 2026-10-26