Your agent. Your engine. Your models.
QGI Studio doesn't lock you to one model vendor. The engine is a per-run choice in the prompt bar, and each engine's model catalog is refreshed live from the fleet: what you see is what resolved, not what a config file says.
Eight engines, live catalogs.
Availability is per tenant; admins enable engines and see their live catalogs under Admin, Engines. Expand a row for the model list.
| Engine | Models (as resolved on the fleet) | Expand |
|---|---|---|
| claude-code | Fable 5 (default) · Opus 5 · Sonnet 5 · Haiku 4.5 | |
| cursor / cursor-cli | QGI-One · QGI-One Fast | |
| Codex | GPT-5.6 Sol · GPT-5.6 Terra · GPT-5.6 Luna · GPT-5.5 | |
| kimi | Kimi K2.7 Code · Kimi K2.7 Code Highspeed · Kimi K3 | |
| opencode | 7 community models | |
| Quark | 45 models, including the Theomax image / video / TTS family | |
| QGI | 48 models: everything in Quark plus QGI-Ultra, QGI-Ultra+, QGI-Ultra+ Fast |
Images, video, speech and vision. Local first.
The media collection routes to the QGI GPU fleet: Theomax Image, Image Layered and Edit for stills; Theomax Lightning and Theomax Pro for text-to-video; Theomax TTS for speech; transcription, OCR, embeddings and a vision model that can judge a rendered page. Cloud models (Theomax 2, Theomax 3 Ultra, Hygen Avatar) are one call away when the local look isn't enough. The tool descriptions even tell the agent which is free and which costs money.
Different engines are good at different things.
Long autonomous builds on claude-code; quick edits on cursor-cli; cheap batch drafting on an open model; media on the local fleet. Choose per run, keep everything else the same: tools, files, history.