Liquid AI adds vision to d1

Liquid AI announced that d1 now handles images, a week after its experimental text-only release, which we covered earlier. Vision is available now through console.liquid.ai and the d1 Playground.

d1 does not generate tokens. It reads text, images or both, along with one or more questions, in a single forward pass and returns a probability for each possible answer. It supports three question types: yes/no, choice from a set of labels, and a score on a scale. Liquid says a text decision takes 200 to 300 ms.

The company compared d1 with GPT-6.1 Sol and Claude Opus 5.5 on six applications. It says d1 matched or beat GPT-6.1 Sol on four of them, cost 19x to 200x less than both models, and answered faster on every task. Liquid ran the tests itself, with one run per model, on October 5, 2026.

On the public VisA inspection dataset, which covers circuit boards, candles, cashews and chewing gum, d1 sorted good parts from defective ones with 85–97% accuracy without task-specific training. In Tetris, adding the screen image raised its score from 70 to 81 cleared lines. In Wordle, it solved all 12 games from screenshots alone.

d1 is billed only on input tokens, at $0.04 per million. Images cost 1.5 tokens per 32×32 patch, so a 1024×1024 image comes to 1,536 tokens. On Vercel and OpenRouter, d1 is text-only for now. Liquid says open weights for upcoming models will come to Hugging Face soon.