Liquid AI releases weights and demos for two multimodal d1 decision models.
Liquid AI releases multimodal d1 decision models for edge devices
2026-10-08Verified announcementTRIAL
Liquid AI now offers both downloadable d1 decision weights and the hosted multimodal d1-ultra API. d1-ultra returns fixed-choice probabilities for text and image decisions at $0.042 per million input tokens with no output-token charge. Vendor evaluations across 112 tasks report gains over Jev 1.14; independent results are not yet available.
What this means for you
For routing, moderation and edge inspection, a compact model can move structured decisions off the generative path and reduce latency. The benchmarks are vendor-run, the custom LFM license needs review and trust_remote_code expands the execution boundary.
Evidence & related material
Liquid AI launches the hosted d1-ultra decision API
Subsequent changes
Liquid AI launches the hosted d1-ultra decision API
Technical details & scores
| Models | d1-3B and experimental d1-omni-600M |
|---|---|
| Outputs | typed noul, choice and number decisions in one forward pass |
| Reported Latency | d1-3B: 8 ms RTX 4090, 9 ms AMD MI325X, 30 ms Apple M5 Pro, 50 ms Jetson Orin Nano |
| Reported Scores | Decision Index 0.2.1 48.57 for d1-3B; seven-dataset mean 82.9 |
| License | LFM 1.0 custom model license |
| Execution Note | requires transformers>=5.14 and trust_remote_code=True |
| Hosted API | d1-ultra fixed-choice probabilities for text and images |
| Pricing | $0.042 per million input tokens; no output-token charge |
Technical impact 4
Engineering value 4
Adoption signal 2
Maturity 2
Verification cost 2
Risk 3