Recursive ships Flint: open-weight local coding model matches Inkling at 1/10th the size
stanfordnlp · x · 2026-10-26
Recursive, Richard Socher's startup, released its first model Flint — a fast local coding model built with their Eureka Machine automated research loop, with open weights on Hugging Face.
Stanford NLP professor Christopher Manning notes it performs just slightly below Thinking Machines' Inkling despite being >10x smaller, joking he'll wait a bit before buying that Mac Studio. Flint is designed for low-latency local coding on unified-memory workstations, a sign that small specialized open-weight coding models are closing the gap fast.
More from coding & agent
- Why custom inference chips are easier than you think: four reasons from a frontier-lab analysis — IgorCarron · 2026-10-26
- Users warn Claude's idle compaction can trigger false cyber safety blocks on all prompts — andersonbcdefg · 2026-10-26
- Arize AI workshop: how to write your first AI eval, from vibes to LLM judges — AI Engineer · 2026-10-26
- Blind test of 5 self-hosted AI agents on Qwen 3.8 27B: one fails by asking permission — GapNew4766 · 2026-10-26
- J-Chain: an imageboard where AI agents post, reply and develop taste in a social space — zer0_state · 2026-10-26
- CLI Agents Are No Longer the Ideal Interface, Says Elvis Omarsar as Grok Bot and Codex App Take Over — omarsar0 · 2026-10-26