Vresk

Vresk is a multi-model chat workspace built on open models. It returns answers, live widgets and designed decks, labels every reply with the model that wrote it, keeps history in one account, and charges a flat $20 a month with no training on user chats.

Vresk is a single chat workspace that sits on top of a curated shelf of open-weight models. Instead of maintaining separate accounts with several AI providers, a user opens one conversation, picks an engine, and asks. Conversation history and the memory the product keeps about the user live in the Vresk account rather than inside any one model, so changing engines does not mean moving or rebuilding anything. The output comes in three shapes, and no mode has to be chosen first. Answers stream into the thread with highlighted code and typeset maths. Live widgets, such as a calculator, a toggleable chart or a scrollable timeline, run in a panel beside the chat. Decks are full presentations: larger models lay out each slide themselves, while the free models write the text and Vresk's own layouts handle the design. Decks can be revised by asking, stepped back through earlier versions, and exported as a PowerPoint file with real text boxes. Written documents download as PDF or Markdown. Transparency is a central design choice. Every reply carries an "Answered by" label naming the exact model, plus a badge showing whether it was built in the US, Europe or China. The shelf currently includes models such as DeepSeek V4 Pro, GLM 5.3, GPT-OSS 120B, Kimi K3, Llama 4 Scout, Mistral Small 3 and several Qwen and Gemma variants. A switch limits the shelf to US- and EU-built models, and the same ask can be run across several engines side by side for comparison. Conversations are stated never to be used to train AI models, and memory can be turned off or wiped. Vresk suits people who want to use different open models for different jobs without juggling subscriptions: analysts building slide decks, students and developers wanting quick answers, and anyone curious about comparing models. It competes with single-vendor chat apps and with multi-model aggregators. Its pitch is flat pricing with a weekly allowance, but billing is limited to the United States and Canada, excluding Quebec.