LLM Releases
← Catalog

SWE-2

Available
CognitionProprietary

Cognition's proprietary coding model for long-horizon, asynchronous software-engineering tasks in Devin (launched Sep 10, 2026). Post-trained from the Kimi K3 2.8T-parameter base with additional reinforcement learning at multi-trillion-parameter scale, SWE-2 targets the cost–performance Pareto frontier for autonomous SWE work. Available at launch through Devin Desktop and CLI, with rollout to Devin Web and Fusion. Proprietary — weights not released; served only via Devin (no public per-token API price).

Specifications

License
Proprietary
Weights
Not released
Architecture
Mixture-of-Experts
Parameters
2.8T
Context window
— tokens
Max output
Knowledge cutoff
Price (in / out, $/M)
Modalities
TextCode

Benchmarks

No benchmark scores recorded yet. Spotted some? Submit a correction.

Vendor-reported figures are claims until independently verified. See methodology.