Skip to content
HN On Hacker News ↗

Introducing SWE-2: Pushing the Pareto Frontier

▲ 446 points 194 comments by seelos 1w ago HN discussion ↗

Pangram verdict · v3.3

We believe this text is mainly human-written, with some AI and AI-assisted content.

6 %

AI likelihood · overall

Human
96% human-written 1% AI-generated
SEGMENTS · HUMAN 1 of 2
SEGMENTS · AI 0 of 2
WORD COUNT 91
PEAK AI % 31% · §2
Analyzed
Sep 10
backend: pangram/v3.3
Segments scanned
2 windows
avg 46 words each
Distribution
96 / 1%
human / AI fraction
Verdict
Human
Pangram v3.3

Article text · 91 words · 2 segments analyzed

Human AI-generated
§1 Human · 13%

By The Cognition Team09.10.26Today we’re introducing SWE-2, our most advanced coding model yet. It pushes the Pareto frontier of capability and cost, achieving 50.0% on FrontierCode 1.1 Main1, within one point of Fable 5.1 while being 64% cheaper.With SWE-2, we scaled RL to the multi-trillion-parameter regime for the first time, building on the SWE-1.72 training infrastructure and recipe.

§2 Mixed · 31%

The key addition is an RL algorithm that trains all reasoning-effort levels in a single run, advancing the whole cost–performance frontier.base modelend of trainingThe result is our closest model yet to the frontier.