Pixel Canary is a stealth-stage coding model distributed through Vercel's AI Gateway under the identifier stealth/pixel-canary, offered free while it remains in preview. Vercel positions it as strong at coding tasks such as building applications and refactoring existing code, and particularly well suited to frontend development and mobile app design, covering responsive layouts, app screens, navigation, and interactive components. Its creator has not been publicly confirmed, though independent reporting notes that tokenizer analysis hints at a possible Qwen lineage, a clue rather than a definitive attribution. The model is also surfaced inside developer coding agents such as Cline, where it is presented as an anonymous large model with strong coding capabilities, broadening its reach beyond a single interface.
On Vercel's Next.js evals, Pixel Canary ties GPT 6 Astra (high) at a 90.3% baseline success rate by passing 28 of 31 tasks, and when Next.js documentation is supplied through an AGENTS.md file, it climbs to 30 of 31 tasks at 96.8%, tying the leaderboard's top score in that setting. Successful tasks span App Router migrations, data fetching, image and font optimization, caching, and view transitions, with scoring measured using pass@4, meaning a task passes if any of up to four attempts succeeds. Practically, the model fits developers working on Next.js and adjacent web or mobile frontends who want a free, capable assistant for code generation and refactoring, while accepting that zero-data-retention is unavailable and that prompts and responses may be used for training and model improvement.
