{"slug": "flint-40b-frontier-local-llm-fast-125-tps-on-macbook", "title": "Flint 40B: Frontier Local LLM (Fast 125 Tps on MacBook)", "summary": "Recursive SI released Flint, its first model, an open-weights local coding model built with the company's automated research system, the Eureka Machine, CEO Richard Socher announced on X. Recursive SI claims Flint runs at 125+ tokens/sec on an M4 Pro MacBook Pro and 150+ tokens/sec on a 64GB NVIDIA RTX Spark, up to 2x faster than comparable models on short contexts and 5x+ faster than dense models with similar coding scores, and that it is the best-performing MoE model under 50B on SWE-bench Pro v2, Terminal-Bench 2.1 and SWE-bench Verified. The model was built from scratch in under 4 months, with weights available on Hugging Face and free trials on OpenCode and Pidotdev.", "body_md": "Richard Socher on X: \"Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.\n• Best-in-class fast local coding\n• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark\n• Built from… / X\n\nRichard Socher on X: \"Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.\n• Best-in-class fast local coding\n• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark\n• Built from scratch in under 4 months\nPrivate, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.\"\n\nIntroducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.\n• Best-in-class fast local coding\n• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark\n• Built from scratch in under 4 months\nPrivate, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.\n\nIntroducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.\n• Best-in-class fast local coding\n• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark\n• Built from scratch in under 4 months\nPrivate, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.\n\nFlint is the best-performing model in its class (MoE models under 50B) on agentic coding benchmarks like SWE-bench Pro v2, Terminal-Bench 2.1, and SWE-bench Verified.\nIt also rivals similarly sized dense models that are much slower to run.\n\nFlint is also the fastest model in its class:\n• 125+ tokens/sec on an M4 Pro MacBook Pro\n• 150+ tokens/sec on a 64GB NVIDIA RTX Spark\nUp to 2x faster than comparable models on short contexts, and 5x+ faster than dense models with similar coding scores. Your data stays on yourShow more\n\nFlint is also a real-world test of the Eureka Machine, our system for automated AI research.\nIt runs the full research loop: ideation, implementation, and validation. Then it uses the results to decide what to try next. We used it in nearly every part of how Flint was built.Show more", "url": "https://wpnews.pro/news/flint-40b-frontier-local-llm-fast-125-tps-on-macbook", "canonical_source": "https://twitter.com/RichardSocher/status/2112944385415426249", "published_at": "2026-10-21 18:57:14+00:00", "updated_at": "2026-10-21 19:13:41.604531+00:00", "lang": "en", "topics": ["large-language-models", "ai-research", "ai-tools", "artificial-intelligence"], "entities": ["Recursive SI", "Flint", "Eureka Machine", "Richard Socher", "Hugging Face", "OpenCode", "Pidotdev", "NVIDIA RTX Spark"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/flint-40b-frontier-local-llm-fast-125-tps-on-macbook", "markdown": "https://wpnews.pro/news/flint-40b-frontier-local-llm-fast-125-tps-on-macbook.md", "text": "https://wpnews.pro/news/flint-40b-frontier-local-llm-fast-125-tps-on-macbook.txt", "jsonld": "https://wpnews.pro/news/flint-40b-frontier-local-llm-fast-125-tps-on-macbook.jsonld"}}