Liquid AI Releases d1-ultra Multimodal Decision Model for $0.042 per 1M Input Tokens

Liquid AI released d1-ultra, a hosted multimodal decision model that processes text and images in a single forward pass with 0 output tokens, priced at $0.042 per 1M input tokens on the Liquid API. According to vendor-reported data, d1-ultra ranks first on the Artificial Analysis Decision Index with a score of 65 and leads all 8 categories across Liquid's 112-evaluation suite, outperforming TypeSafe's Jev 1.14 at the same price point. Instead of generating free text, the model returns bounded probabilities for yes/no questions, label choices, or ordered rubric scores to handle routing, moderation, and LLM-as-judge workflows. Billing applies exclusively to input tokens at 1.5 tokens per 32x32-pixel image patch (1,536 tokens for a 1024x1024 image), but the API bills each question in a request as a separate prompt—meaning 3 questions on 1 image bills that image 3 times. Liquid AI built d1-ultra on a larger customized open-weight backbone than its earlier d1 model, though exact parameter counts, context length, downloadable weights, and per-benchmark breakdowns remain undisclosed.