Richard Socher's Recursive debuts Flint, an AI-built local coding model hitting 125 tokens/sec on a MacBook
dr_cintas · x · 2026-10-22
Recursive, founded by Richard Socher, released Flint, a fast local coding model built almost end-to-end by its automated research system, the Eureka Machine, in under four months with a handful of researchers. WIRED frames it as a milestone for recursive self-improvement: the Eureka Machine devised new base-model training methods, architectures, and chip-specific optimized code.
- Speed: 125+ tokens/sec on MacBook Pro, 150+ on RTX Spark — up to 2x faster than comparable models; open weights on Hugging Face
- Benchmarks (40B total/2B active): Terminal-Bench 2.1 54.9, SWE-bench Pro v2 89.7, SWE-bench Verified 74.2, LiveCodeBench v6 80.3, leading similar-hardware local models
- Pitch: private, unmetered on-device coding intelligence
More from coding & agent
- Migrating an entire Hermes automation workload to Antigravity took about 2 hours — rseroter · 2026-10-22
- How a second Mac mini lets AI agents auto-distribute compute-heavy tasks — tedddyoweh · 2026-10-22
- LangChain's Design Agent Shipped 3,000+ Assets and Earned 4.5M Views Since August — LangChain · 2026-10-22
- Databricks Chief AI Scientist: Evals Are the Biggest Bottleneck to Enterprise AI Adoption — jefrankle · 2026-10-22
- Developer plans a solar-aware job scheduler so his homelab rides free excess power — mertdumenci · 2026-10-22
- How do you design your agentic code review workflow? — lucasmeijer · 2026-10-22