Understand this story
On October 21, 2026, Recursive Superintelligence released Flint, an open-weight local coding model. According to related reports, Flint is a 40B mixture-of-experts model that activates approximately 2.5B parameters per token, is licensed under Apache 2.0, and its weights have been made available on Hugging Face; it was built from scratch in under 4 months by the automated research system Eureka Machine. On October 22, Hugging Face co-founder Thomas Wolf commented that Recursive researchers quantized the lm_head—which is usually recommended to be kept at high precision—to 4 bits, and the result differed from bf16 by only 0.7 percentage points. On the same day, OpenRouter launched Flint, saying it is aimed at coding agents, supports a 262K context, reasoning, and tool calling, and is free to use during the launch period; Unsloth released a Dynamic GGUF quantized version of Flint, saying it can run locally on devices with 24GB of memory. The model now offers multiple usage paths through open weights, hosting platforms, and quantized versions.
AI-generated from public reports
Latest development · Report
Edge AI Daily Morning Brief: OpenAI opens decision API, Google releases ACEDevelopments
4 confirmed developments
Thomas Wolf commented on an AI researcher quantizing lm_head to 4 bits
Source:Thomas WolfUnsloth offers a local running solution for Flint's Dynamic GGUF
Source:Unsloth- Source:OpenRouter
- Source:36Kr · AI
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 24
TMTPost · AGI
Edge AI Daily Morning Brief: OpenAI opens decision API, Google releases ACE
OpenAI fully opens the decision API based on GPT-6 Luna, with input at only $0.1 per million tokens and output tokens free, and response speed up to 10 times faster than calling the same model via the Responses API. Google releases ACE, a dedicated compute engine for AI agents, allocating a hardware-isolated microVM to each agent and achieving 500 activations per second under a benchmark of 200,000 agents. Anthropic donates the Agent Skills open standard to the Agentic AI Foundation under the Linux Foundation.
Oct 23
36Kr · AI
Mirendil and Recursive release RSI results, NVIDIA researcher discusses survival rules for researchers
Startup Mirendil claims that one researcher, with the help of AI agents, wrote over 99% of the code and trained a 100B–500B parameter coding foundation model, Mirendil-Base, in 3.5 months with under $10 million in compute; Tian Yuandong's startup Recursive released Flint, an open-weight 40B coding model, backed by the Eureka Machine system that automatically generates ideas, implements, tests, and decides the next step. NVIDIA senior researcher Ali Hatamizadeh wrote about this, arguing that conventional pretraining recipes are now mature and public, and that researchers' value has shifted to deciding which experiments to run and judging which results are trustworthy; he lists seven open topics: few-shot efficient learning, new architectures, continual learning, long-horizon reasoning reliability, eliminating hallucinations, new evaluation systems, and interpretability and control.
Oct 22
@mark_kRelated discussion
@mark_k joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
The Decoder
Recursive Releases Its First Model Flint, Developed with Participation from Eureka Machine
AI startup Recursive has released its first language model, Flint, with its Eureka Machine system involved in almost every step of the R&D process, as the system itself proposed ideas, implemented them, tested them, and planned subsequent experiments. Flint is a MoE model for agentic coding on local hardware, with 40B total parameters and 2.5B activated per token. It was completed in four months, and its weights are released on Hugging Face under the Apache 2.0 license, supporting llama.cpp, Ollama, LM Studio, vLLM, and SGLang. Recursive claims that it leads MoE models with fewer than 50B parameters on SWE-Bench Pro v2, Terminal-Bench 2.1, and SWE-Bench Verified, but the results have not yet been independently verified.
36Kr · AISelected
Recursive open-sources 40B MoE coding model Flint, runnable on MacBook Pro
Recursive, founded by Tian Yuandong, has open-sourced its first model Flint, focused on local Agent coding, with 40 billion total parameters, 2.5 billion activated per token, and support for a 512k token context window. The company says that paired with its self-developed inference engine rec, Flint can reach at least 125 tokens/s on a MacBook Pro and over 150 tokens/s on RTX Spark, roughly 1.4–2.6 times higher throughput than llama.cpp. Flint scores 74.2% on SWE-bench Verified, 89.7% on SWE-bench Pro v2, and 54.9% on Terminal-Bench 2.1; rec currently supports M4 Pro, M5 Pro, and M5 Max, requiring at least 48GB of unified memory and about 30GB of free disk space.
Zhidx
Recursive, co-founded by Yuandong Tian, open-sources Flint, a 40B local coding model
Recursive Superintelligence, co-founded by former Meta FAIR research director Yuandong Tian, has launched its first local coding model, Flint, with 39.8 billion total parameters and approximately 2.5 billion activated parameters per token. It has been open-sourced under the Apache 2.0 license, with weights available via Hugging Face. The model generates over 150 tokens per second on NVIDIA RTX Spark and over 125 tokens per second on a MacBook Pro with M4 Pro, and scores 54.9%, 89.7%, and 74.2% on Terminal-Bench 2.1, SWE-Bench Pro v2, and SWE-Bench Verified, respectively. Recursive says its team completed the entire training pipeline from scratch in under 4 months with the help of its automated research system Eureka Machine.
Shao Meng
Recursive releases its first model Flint: local coding MoE, 125+ tokens/s on a MacBook
Recursive releases its first model Flint, a MoE with 40 billion total parameters and 2.5 billion activated parameters per token, aimed at local Agent coding, with weights open-sourced under Apache 2.0. With short context, it exceeds 125 tokens per second on an M4 Pro MacBook Pro and over 150 on a 64GB RTX Spark, and supports llama.cpp, Ollama, LM Studio, vLLM, SGLang, and Unsloth. Nearly every stage of its R&D involved the automated research system Eureka Machine, including architecture search, quantizing the output layer to 4 bit with offline distillation, and fusing Metal kernels for M4/M5.
UnslothFirst-partySelected
Unsloth releases the Dynamic GGUF quantized version of Flint, runnable locally on 24GB of memory
Unsloth has released the Dynamic GGUF quantized version of Flint, which can run locally on 24GB-memory devices, along with a running guide. Flint is Recursive's Apache-2.0 model with 39.8B parameters and 2.5B active, which Unsloth calls the strongest MoE model under 50B parameters, reaching 125 t/s on Mac and 150 t/s on RTX Spark. Recursive co-founder Richard Socher previously introduced that the model was built from scratch in under 4 months by the automated research system Eureka Machine, and the weights are now open on Hugging Face.
MarkTechPost
Recursive Releases Open-Weight Coding Model Flint, a 40B MoE with Only 2.5B Active Parameters
Recursive Superintelligence has released its first model, Flint, a 40B MoE model aimed at coding agents with roughly 2.5B active parameters per token; the weights are released under the Apache 2.0 license on Hugging Face. The model card lists a context of 262,144 tokens, with 4-bit weights at about 23 GB, runnable on a 48 GB Apple Silicon Mac or a 48 GB NVIDIA GPU. The company self-reports SWE-bench Verified at 74.2% and IFEval at 95.1%, while Terminal-Bench 2.1 stands at 54.9%, below dense Qwen 3.8's 73.0%. Recursive says its automated research system Eureka Machine contributed to the design of the architecture, quantization, and kernels.
OpenRouterFirst-party
Recursive's first open-source model Flint launches on OpenRouter, a 40B MoE built specifically for coding agents
Recursive's first model Flint has launched on OpenRouter. It is an open-weight 40B MoE model (about 2.5B active parameters), built by the company's automated research system Eureka Machine, aimed at coding agents, and supports 262K context, reasoning, and tool calling. Recursive reports that it scores 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2, and 74.2% on SWE-bench Verified, and is free to use during its launch period.
TechCrunch AI
Recursive releases laptop-class coding model Flint, says it was built mainly with AI involvement
AI startup Recursive has released its first model, Flint, a coding model that runs offline on a laptop and is freely downloadable on Hugging Face under the Apache 2.0 license; the company says its development process was mainly carried out with the involvement of its in-house Eureka Machine. Flint is an MoE model that activates only 2.5B of its 40B parameters at a time, generates at about 131 tokens/sec on a MacBook Pro with an M4 Pro, and outperforms other models under 50B parameters on SWE-Bench Verified. Recursive says Flint requires more guidance than large paid coding assistants and recommends breaking large tasks into small steps.
Testing Catalog
Recursive Releases Flint, Its First Open-Weight Coding Model
Recursive has released Flint, its first open-weight coding model, under the Apache 2.0 license. It is a 40B-parameter MoE model aimed at local coding scenarios. Recursive says it reaches over 220 tokens per second on a 64GB RTX Spark, 2x faster than comparable models. Its automated research system, Eureka Machine, also generated custom GPU code that sped up Flint on the RTX Spark by 55% without changing the weights.
Rohan PaulSelected
Recursive releases Flint, a 40B MoE coding model, at 125+ tokens/s on a MacBook Pro
Recursive has released its first model, Flint, a 40B mixture-of-experts coding model under the Apache 2.0 license, with weights now open on Hugging Face, built from scratch in under 4 months by the automated research system Eureka Machine. Flint generates code at 125+ tokens/s on a MacBook Pro and reaches 150+ tokens/s on an RTX Spark. On a 48GB MacBook, its average quality is about 81%, compared with about 85% for Qwen3.8 27B, but in 128K-token long sessions it is about 7 times faster than the latter.
Thomas Wolf
Thomas Wolf comments on Recursive's new model Flint: quantizing lm_head to 4 bits still comes close to bf16
Hugging Face co-founder Thomas Wolf said that his AGI standard remains AI winning a Nobel Prize, but the AI researchers at Recursive quantized lm_head—which is usually advised to be kept at high precision—to 4 bits, and the result differs from bf16 by only 0.7 percentage points, which impressed him. He also mentioned that the model was released as open weights on Hugging Face, and congratulated Recursive. The quoted original Recursive post introduced that Flint is its first model, positioned as a fast local coding model, built by the automated research system Eureka Machine, reaching 125+ tokens/sec on a MacBook Pro and 150+ on RTX Spark, and built from scratch in under 4 months.
Wired AI
Recursive Releases Open-Source Coding Model Flint, Says Most of Development Process Was Completed by AI
Recursive released Flint, its first open-weight coding model, with 40B parameters. CEO Richard Socher said that nearly every stage of the model pipeline was completed by its Eureka Machine system, including a new base training method, the model architecture, and code optimized for specific chips. A WIRED author tested it on a Mac with 48GB of memory, using it to build a command-line to-do app, a web research scraping program, and an RSI educational website, and found its speed and accuracy superior to models of the same size. Recursive has raised $650 million from Google Ventures, Greycroft, Nvidia, and AMD Ventures, at a valuation of $4.65 billion.
Heat trend for this story
- Current
- 5
- Period peak
- 92 Oct 23 00:00
- 24-hour change
- -90%
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.