Liquid AI (@liquidai) on X
Short summary
Liquid AI released d1 with image support, a decision model designed for yes/no and classification tasks. It matches or beats GPT-6.1 Sol on most benchmarks while costing 19-200x less and running 200-300ms per decision.
- d1 processes images and text to output probabilities for binary, choice, or scoring questions in a single forward pass without token generation
- Benchmarked against GPT-6.1 Sol and Claude Opus 5.5 on six real applications (support ticket filtering, circuit board inspection, etc.), matching or beating GPT-6.1 Sol on four of them
- Costs 19-200x less than comparable frontier models with significantly faster inference speed
Full summary
Liquid AI rolled out d1 with vision capabilities, expanding their decision model to handle images, text, or both together. They tested it against GPT-6.1 Sol and Claude Opus 5.5 on six real-world applications—support ticket filtering, circuit board inspection, and others. d1 matched or beat GPT-6.1 Sol on four of those tasks while costing 19 to 200 times less and running faster across every single one.
The model's design focuses on speed and efficiency over token generation. Instead of producing text output, d1 returns probabilities for yes/no questions, multiple choice, or numeric scores. It completes decisions in 200 to 300 milliseconds with a single forward pass, which explains the dramatic cost and latency advantages. This approach trades flexibility for raw performance on bounded decision problems.
The practical angle here matters. Most AI spend goes toward models that generate lengthy text responses, but many business problems just need a quick classification or ranking. If d1 actually delivers on those numbers—particularly the cost gap—it addresses a real inefficiency in how companies currently deploy AI for routine decision-making tasks.