← AI Switchboard
MODELS · November 11, 2026
Nov 4

Cognition releases SWE-2 Flash, a coding model built on Thinking Machines' Inkling-Small, and keeps the weights back over cyber risk

By the AI Switchboard · researched and written with AI from the sources listed on this page, each point checked against its source before publishing · how we work · report a mistake

Cognition says it runs at a fifth of SWE-2's price and three times its speed, and it is free in Devin until 30 November.

Cognition, which makes the Devin coding agent, released SWE-2 Flash on 4 November. “We began with a large-scale mid-training stage on top of Inkling-Small, a model with 276B total and 12B active parameters,” its post says. Inkling-Small is a Thinking Machines model, not the larger Inkling 2 Small that the Switchboard reported on 7 November.

Cognition reports that the extra training lifted Terminal-Bench 4.0 from 0.9% for the base model to 20.0%. It says Flash runs “at one-fifth the price and 3x the speed of its much larger sibling, SWE-2”.

It is available in Devin Desktop and the Devin CLI, free until 30 November. Cognition is not sharing the weights publicly yet because of cyber risk, and says it is building a way to share them securely with trusted partners and research labs.

Holding back weights over cyber capability is becoming common among smaller labs as well as large ones. The benchmark figures are Cognition's own.

  • Confirmed SWE-2 Flash was trained on top of Inkling-Small, a model with 276B total and 12B active parameters. Cognition
  • Claimed One-fifth the price and three times the speed of SWE-2; Terminal-Bench 4.0 rises from 0.9% to 20.0%. Cognition
  • Confirmed The weights are not being shared publicly yet; it is free in Devin until 30 November. Cognition
Sources: Cognition

Models & releasesSafety, security & governance

Catching up in the November 11, 2026 edition · front page