Spark

Apache

<p><strong>Recursive <a href="https://recursive.com/updates/introducing-recursive-flint" rel="noopener noreferrer nofollow">launched Flint</a> on October 21, 2026. Much of the research behind this open-weight coding model was conducted with the help of AI. The entire model was created with the assistance of a small team in less than four months. The question is whether this approach creates a chance for small laboratories to compete with big players without huge investment.</strong></p>
<p>Flint uses a mixture-of-experts design. It activates only a small part of the model at a time, helping it run efficiently. Its <a href="https://huggingface.co/RecursiveSI/flint" rel="noopener noreferrer nofollow">model card</a> lists 39.8 billion parameters, with about 2.5 billion active per token. The Apache 2.0 license allows commercial use, and developers can run Flint locally.</p>
<h3 class="wp-block-heading" id="strong-what-recursive-says-the-eureka-machine-built-strong"><strong>What Recursive says the Eureka Machine built</strong></h3>
<p>Recursive credits its Eureka Machine with suggesting, coding and testing changes to Flint’s architecture, training and inference software. CEO Richard Socher told <a href="https://www.wired.com/story/recursive-self-improvement-flint-latest-model-scary-good/" rel="noopener noreferrer nofollow">WIRED</a> that AI improved “nearly every aspect of the entire model pipeline.”</p>
<p>That is central to Recursive’s goal. It wants AI systems to help build better successors. Socher maintains that people can still stop the research process.</p>
<h3 class="wp-block-heading" id="strong-benchmarks-from-the-company-a-hands-on-from-wired-strong"><strong>Benchmarks from the company, a hands-on from WIRED</strong></h3>
<p>In <a href="https://recursive.com/updates/introducing-recursive-flint" rel="noopener noreferrer nofollow">tests conducted by Recursive</a>, Flint outperformed other mixture-of-experts models having less than 50 billion parameters on SWE-Bench Pro v2, Terminal-Bench 2.1 and SWE-Bench Verified. The company also reported speeds above 125 tokens per second on an M4 Pro MacBook Pro and 150 on an NVIDIA RTX Spark at short context lengths.</p>
<div class="wp-block-image">
 <figure class="aligncenter size-large">
  <figure style="width: 100%" class="wp-caption alignnone"><img src="https://www.cryptopolitan.com/wp-content/uploads/2026/10/flint-vs-rivals-editorial-v4-4k-1024x576.png" alt="" width="1024" height="576">
   <figcaption>
    Recursive Flint vs Rival AI Coding Models: Coding Benchmarks, Speed and Parameters
   </figcaption>
  </figure>
 </figure>
</div>
<p>One important consideration: these are company numbers. Independent evaluation is still limited. WIRED’s Will Knight tested Flint himself and found it to be quite adept, even able to create a small website. However, some successful coding tasks do not indicate how it compares with other models.</p>
<p><a href="https://hub.openhands.dev/blog/recursive-flint-on-openhands" rel="noopener noreferrer nofollow">OpenHands conducted tests</a> on Flint prior to its launch and provided developers access to it on the day of its roll-out. Although this indicates Flint is capable of performing real-world coding tasks, OpenHands hasn’t conducted an independent assessment of Recursive’s claims regarding their benchmarking.</p>
<h3 class="wp-block-heading" id="strong-the-410-million-question-about-cheap-ai-strong"><strong>The $410 million question about cheap AI</strong></h3>
<p>Creating a model that requires fewer people is not a guarantee that it will be cheap. Recursive has not revealed how much it costs to build Flint.</p>
<p>On July 28, 2026, the company entered in a <a href="https://techcrunch.com/2026/07/28/recursive-superintelligence-signs-400-compute-deal-with-amazon/" rel="noopener noreferrer nofollow">$410 million deal with AWS</a>. Previously it raised $650 million. And none of these amounts show the cost of building Flint itself.</p>
<p><a href="https://epoch.ai/gradient-updates/r-and-d-vs-training-compute" rel="noopener noreferrer nofollow">Epoch AI’s research</a> indicates why expenses stay high. Final training runs made up 9.6% of OpenAI’s research-compute expenditures during the year 2024. The majority of the funding was distributed to other research-related actions.</p>
<p>AI agents can take over some jobs that used to be handled by researchers. However, running experiments and testing new ideas still require a big amount of computing power.</p>
<h3 class="wp-block-heading" id="strong-open-weights-local-hardware-and-global-competition-strong"><strong>Open weights, local hardware, and global competition</strong></h3>
<p>Meanwhile, using AI is becoming much cheaper. <a href="https://epoch.ai/publications/the-plunging-price-of-thought" rel="noopener noreferrer nofollow">Epoch AI estimates</a> that the cost of achieving comparable performance has fallen about 47% per quarter since 2023.</p>
<p>On September 22, 2026, OpenAI introduced GPT-6 Sol and Luna at prices at least 50% below their predecessors, as <a href="https://www.cryptopolitan.com/openai-cuts-gpt-6-sol-luna-prices-by-50/" rel="noopener noreferrer nofollow">Cryptopolitan reported</a>.</p>
<p>Flint had nothing to do with those price changes. Still, it gives developers another option: a coding model they can run on their own hardware without relying on paid cloud services.</p>
<p>The benefits could reach beyond established AI markets. <a href="https://hai.stanford.edu/ai-index/2026-ai-index-report" rel="noopener noreferrer nofollow">Stanford’s 2026 AI Index</a> found that open-source development is helping more countries participate in AI innovation.</p>
<p>For now, Flint shows what a small team can accomplish with AI-assisted research. Whether the same approach can make model development affordable for smaller labs remains an open question.</p>
<p>Don’t just read crypto news. Understand it. Subscribe to our newsletter. <a href="https://www.cryptopolitan.com/newsletters/?utm_source=cp&amp;utm_medium=web&amp;utm_campaign=inlineAds" rel="noopener noreferrer nofollow">It's free</a>.</p>

Recursive launches Flint coding model built largely with AI

Table of Contents

Cryptopolitan_News

Other articles published on Oct 22, 2026

Join our free newsletter for daily crypto updates!