## v1.2.012 — 2026-06-20
- fix: **Importing a local GGUF model no longer reports "architecture not supported" for newer model families — such as Gemma 4 — that the built-in engine can actually run.** Imports were checked against a fixed list of known architectures that lagged behind the bundled inference engine, so a model that loads fine elsewhere (for example in LM Studio) could be rejected here even though it would run. The check is now inverted: any architecture the engine can load is accepted, and only non-chat model types (text-embedding models, vision encoders, and multimodal projector files) are declined. Newer chat-model architectures now import and run without waiting for an app update.
- News and updates
- AI Desktop Pet
- v1.2.012 — 2026-06-20
Source: Steam Community (opens in a new tab)
