d1 is Liquid AI's decision model family. OpenRouter lists d1 and d1.1 at $0.02 per million input tokens, with free output, on 28 Oct 2026.
By Liquid AI · Released 29 Sep 2026
Liquid AI's decision model family. The hosted d1 takes the same three question types as Jev (Noul, Choice and Score). On 7 Oct Liquid released two small open siblings you can run on a device, and on 15 Oct a bigger hosted one, d1-ultra, joined OpenRouter. On 27 Oct d1.1, the successor to d1, joined OpenRouter at the same price, with -latest aliases that follow each family's newest version. On 27 Oct Liquid announced two more open models, d1-2B (text, pictures and video) and d1-Omni-2B (adds audio), built on LFM3; their Hugging Face repositories went up on 24 Oct.
d1: Liquid AI's decision model family.
d1: Liquid AI
d1: Open weights: d1-3B and d1-omni-600M (7 Oct), then d1-2B and d1-Omni-2B (announced 27 Oct), under the LFM Open License v1.0
A community ARM64 assembly engine runs d1-2B Q4_K_M with its Q8_0 projector on a 4 GB Raspberry Pi 4B. It uses a 512-position context and different image-position policies for choices and counts. The maker’s timings are tens of seconds per image question.
Pi setup and timing boundaries · Assembly engine and run records
Source: Docs → Source: Announcement → Source: Decision Index → Source: Open d1 release → Source: d1.1 on OpenRouter → Source: d1-2B →
A pixel-art racer where d1 picks left, centre or right on every tick, with a mode that races it against Jev. Liquid AI's own example, in their cookbook repo, so the star count is for the whole repo.
A script that asks true/false and multiple-choice questions about an image, PDF, Word file, slide deck, spreadsheet or text, using a decision model on local Ollama. Validated with d1:3b and Clef-Flash; d1:3b is the default.
A ComfyUI workflow that adds you to an existing video. Liquid's d1-2B judges every shot (framing, whether there's room for one more person, where everyone stands) and casts the best ones, then image and video models put you in beside the people already there.
Uses Liquid's open d1-2B as a local System 1 for DeepSeek Harness Browser Use and macOS Computer Use, with approvals and routing. Downscaling wide screenshots before the model reads them cut one decision from 879 ms to 310 ms on a 2508-pixel window, and six consecutive trials all chose a click action. d1-2B itself is under the LFM Open License, not MIT.
Supplies OpenClaw with local typed decisions from LiquidAI/d1-Omni-2B-MLX-8bit on Apple Silicon. It has a native decision provider and a bridge for OpenClaw's TypeSafe integration. The repository name contains Kev, but the runner loads d1-Omni. VENDOR-REPORTED on an M4 Mac mini: about 200 ms per warm batch and about 3 GB of resident RAM.
Native Swift and MLX inference for local d1 text, images and speech, with a shared HTTP server and an older von backend. The author publishes Python reference vectors and layer-by-layer parity measurements. Head-logit differences are 5.8e-5 absolute for d1 and 6.0e-5 for von. These are implementation checks, without a task-accuracy benchmark. The HTTP wire accepts text states; native Swift calls carry media.
Author-run reference vectors and source checked 6 Nov 2026; not rerun locally.
Source: Licence file → Source: d1 terms → Source: Integration code → Source: Reference vectors → Source: Port verification →
Moves llama.cpp decision serving code into a C library. Its recorded d1 request takes 678 ms on eight CPU threads. Saved CLI/server parity tests stay within 0.01 for Laya, exceed that probability tolerance for H2O and Clef, and reject the tested Perplexity package. On a Pixel 8a in Termux, the maker reports a 464 ms yes/no decision with d1-2B Q4_K_M on four threads and 2.1 GB RSS. Without repacking, that question takes 1,210 ms at 1.1 GB RSS. iOS and Capacitor app bindings remain plans.
Source: Licence → Source: Saved parity output → Source: Phone measurement and build notes →
ARM64 assembly inference for d1-2B Q4_K_M and its Q8_0 projector on a 4 GB Pi 4B. Source, manifests and 80 answers are public. Choice questions update 48 selected image positions; direct counts keep all 300. Same-input timing uses four Pi pairs per format; the complete reference accuracy run uses x86 llama.cpp.
Author-run records; not independently reproduced
Source: Licence → Source: Benchmark protocol and results → Source: Third-party terms → Source: Image-question helper →