unknown

We are sharing an early preview of our ongoing SWE-1.6 training run.

It significantly improves upon SWE-1.5 while being post-trained on the same pre-trained model - and it runs equally as fast at 950 tok/s. On SWE-Bench Pro it exceeds top open-source models.

The preview model still exhibits some undesirable behaviors like overthinking and excessive self-verification, which we aim to improve. We are rolling out early access to a small subset of users in Windsurf.

Swyx offers insights into software engineering, developer productivity, and career growth, providing articles, talks, and resources to help developers level up their skills and advance their careers in tech. By exploring Swyx's curated content, developers can learn about coding strategies, learning methodologies, and networking techniques for becoming more effective and successful in their profession. Whether you're navigating your first job search, mastering new technologies, or aiming for leadership roles, Swyx offers resources to support your personal and professional growth in the software industry.

swyx

Cognition (the company behind Devin) is developing a custom model called swe-x, positioned as an Opus-tier coding model running at ~1000 tokens/second. The author, initially skeptical about Cognition investing heavily in custom model training, has changed their view after seeing the Cognition for Governments business scaling to billions in ARR. They now see owning a proprietary model—built on being the first and largest cloud coding agents business—as an offensive strategic move rather than a defensive one against larger labs.

# unconstrained Opus-tier koding at ~1000tok/s

if you read The Agent Labs Thesis, custom model training is the smallest part of what an ideal Agent Lab should do. i was initially a skeptic that cog should invest a lot more in swe-x other than as a defensive manuever - cant out-bitter-lesson the big labs, but for “most” or routing tasks you could still do it.

I’ve genuinely been turned around on this. the Cognition for Governments business is scaling to billions in ARR. and having own model based on being the first and largest cloud coding agents biz is offense, not defense.

<p># unconstrained Opus-tier koding at ~1000tok/s

if you read The Agent Labs Thesis, custom model training is the smallest part of what an ideal Agent Lab should do. i was initially a skeptic that cog should invest a lot more in swe-x other than as a defensive manuever - cant out-bitter-lesson the big labs, but for “most” or routing tasks you could still do it.

I’ve genuinely been turned around on this. the Cognition for Governments business is scaling to billions in ARR. and having own model based on being the first and largest cloud coding agents biz is offense, not defense.</p>