Vinci model classes
Products choose a stable capability class, while the model and approved provider can change after evaluation.
Vinci Mezzo
Fast and light for everyday chat. For code and file tasks, Forte works best.
- Current model
- DeepSeek V4 Flash
- Modality
- Text input and output
- Provider
- DeepInfra
- Data handling
- Zero data retention enforced
- Infrastructure
- Canadian- and US-based infrastructure
- Evaluation
- v1
Vinci ForteDefault
Our default. Strong across code, files, and long tasks.
- Current model
- GLM 5.2
- Modality
- Text input and output
- Provider
- DeepInfra
- Data handling
- Zero data retention enforced
- Infrastructure
- Canadian- and US-based infrastructure
- Evaluation
- v1
Vinci Fortissimo
Our deepest reasoning, for the hardest problems.
- Current model
- Kimi K3
- Modality
- Text input and output
- Provider
- Fireworks
- Data handling
- Zero data retention enforced
- Infrastructure
- Canadian- and US-based infrastructure
- Evaluation
- v1
How image attachments work
The chat classes above are text-only today. When you attach an image, Qwen3-VL 30B-A3B on DeepInfra describes it first under the same zero-data-retention rules. Your selected class receives that description, not the raw image. The image and description are not retained.