@devindesktop: SWE-1.7 is now available in Devin Desktop and Devin CLI. It achieves near-frontier performance at a fraction of the cos…
Summary
Cognition introduces SWE-1.7, a cost-efficient model achieving near-frontier performance at 1000 tok/s, available in Devin Desktop and Devin CLI.
View Cached Full Text
Cached at: 07/09/26, 11:36 AM
SWE-1.7 is now available in Devin Desktop and Devin CLI.
It achieves near-frontier performance at a fraction of the cost.
And you can use lightning mode at 1000 tok/s!
Cognition (@cognition): Introducing SWE-1.7, the most capable model we’ve trained yet.
It scores within a few points of the strongest frontier models at a fraction of the cost, and is now available at 1000 tok/s.
RL is not hitting its limit: after refining our recipe, we keep seeing gains as we scale
Similar Articles
@Kyler_Lorin: SWE 1.6 Fast in Devin CLI is actually insane
SWE 1.6 Fast for Devin CLI has been released, promising significant speed improvements for software engineering tasks.
SWE-1.7: Frontier Intelligence at a Fraction of the Cost (22 minute read)
Cognition launches SWE-1.7, a highly capable AI model for agentic software engineering that achieves frontier-level performance at reduced cost, with improvements in RL training, multi-cluster infrastructure, data curation, and self-compaction for long tasks.
@cognition: Claude Sonnet 5 is now available in Devin Desktop and Devin CLI. Sonnet 5 pairs frontier-level coding performance with …
Claude Sonnet 5, a new AI model with frontier-level coding performance at a more affordable price, is now available in Devin Desktop and Devin CLI, outperforming Opus 4.8 on FrontierCode Extended.
@morganlinton: Starting to test out Devin for Terminal, and comparing to Codex. Using Devin's model to start, SWE-1.6 Fast. And holy m…
A developer begins testing Devin for Terminal using the SWE-1.6 Fast model, comparing it to Codex and noting impressive speed during initial trials.
Devin Fusion (8 minute read)
Cognition introduces Devin Fusion, a multi-model harness that routes between frontier and cost-effective models using a sidekick architecture, achieving frontier-level performance at 35% lower cost.