The $48b agent company just proved model value moved to post-training. cognition's swe-2, shipped…
the $48b agent company just proved model value moved to post-training. cognition's swe-2, shipped sept 10, post-trains moonshot's open kim k3 and lands within a point of claude fable 5.1 on frontiercode while costing 64% less.
the frontier is an open base plus the rl layer on top.
Context
Cognition's SWE-2 post, dated 10 September 2026, reports SWE-2 at 50.0% on FrontierCode 1.1 Main, within one point of Fable 5.1 at 50.9% while 64% cheaper, post-trained from Kimi K3, a 2.8T-parameter model that had already had extensive RL for agentic coding. Its table shows FrontierCode 1.1 Main at 50.0 for SWE-2, 44.2 for Kimi K3, 50.9 for Fable 5.1 and 53.3 for GPT-6 Astra, DeepSWE 1.1 at 73.0, 68.5, 67.4 and 74.1, and Terminal-Bench 4 at 27.3 for SWE-2 against 55.8 for Fable 5.1 and 57.9 for Astra. It is available in Devin Desktop and CLI.
These are Cognition's own results on its own runs, three runs per task per model, not independent, and the cost basis behind 64% was not read in full. Within a point holds on FrontierCode 1.1 Main only: SWE-2 is 3.3 points behind Astra there and far behind Fable 5.1 and Astra on Terminal-Bench 4, and scores are not merged across benchmarks. Cognition's page names Kimi K3 and does not state Moonshot's open-weights or license status, which was not read here; the earlier Moonshot note quotes a license with incomplete terms. The 48 billion dollars is TechCrunch's figure and is not proof of where model value sits. In this evidence RL added about 5.8 points over the base on FrontierCode, one vendor data point, so the proof that value moved to post-training is the author's conclusion.
Related work
- Earlier note on Moonshot's K3 ↗Kimi K3, Moonshot.
- Earlier note on the Cognition round ↗The $48B round.
Watch next
- Independent FrontierCode and DeepSWE runs and Cognition's cost methodology.
Sources
- SWE-2 (Cognition, 10 Sep 2026)cognition.com
- Cognition hits 48B valuation (TechCrunch, 8 Sep 2026)techcrunch.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 20 September 2026 at 03:06 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →