Is d1.2 state of the art?Pending benchmarks.Announced Nov 16, 2026; no independent results yet.
Liquid AI's hosted decision model, kept separate from d1.2-ultra. Liquid describes d1.2 and d1.2-ultra as the new versions of its decision models. They take text, images, and audio as input — the audio can be what you ask about, or a spoken question — and answer typed questions in a single forward pass with no generated tokens. New answer types include reading a value (text, number, or date), finding passages, locating objects in a picture, estimating a number with a range, and picking every option that applies. Announced November 16, 2026 (Liquid blog, datePublished 2026-11-16; @liquidai at 2026-11-16T15:08:30Z, 10:08 AM ET). Live on the Liquid API as a self-serve model (id d1.2), billed on input tokens only at $0.02 per million. No public weight release is stated.
The details so far
- Lab
- Liquid AI
- Announced
- Nov 16, 2026
- API
- Live · d1.2
- Open weights
- Closed · No public weight release is stated on the official blog or pricing page. Official Hugging Face pages for LiquidAI/d1.2 and LiquidAI/d1.2-ultra return 404
- Input
- Text, images, and audio. The audio can be what you ask about, or a spoken question (Liquid blog)
- Output
- Typed answers with probabilities; no generated tokens. New types: read a value (text, number, or date), find passages (spans), locate objects in a picture (box / boxes), estimate a number with a range, and pick every option that applies (Liquid blog)
- Hosted access
- Self-serve Liquid API, model d1.2. Create a key at console.liquid.ai (Dashboard > API Keys). Endpoint https://api.liquid.ai/decisions/v1/systemone. New organizations start with $2 in free credit (Liquid blog and pricing page). OpenRouter and Vercel are described as coming soon, so they are not listed as live hosts
- Price
- $0.02 per 1M input tokens; $0.00 output. Billed on input tokens only. An image costs 1.5 tokens per 32×32-pixel patch at the text rate, so a 1024×1024 image costs 1,536 tokens. A request bills its text and pictures once, and each extra question adds only its own text (Liquid blog and pricing page)
- Related
- Separate from d1.2-ultra, and from earlier Liquid decision models d1.1, d1-ultra, d1-2B, d1-3B, d1-Omni-2B, and d1-omni-600M
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: self-reported on Liquid's comparison across 13 benchmarks and 22,505 questions, d1.2's calibration error is 1.2 points, compared with 0.8 for d1.2-ultra, 2.7 for Jev 1.14, 3.7 for Perplexity's Decider v1.3, and 7.2 for GPT-6 Luna on its Decisions API (Liquid blog).
Sources
LABLiquid AI blog: d1.2 and d1.2-ultraliquid.ai/blog/d1-2 · Nov 16, 2026LABLiquid AI pricingliquid.ai/pricing · Nov 16, 2026LABLiquid AI announcement on Xx.com/liquidai/status/2122346077231923698 · Nov 16, 2026LABLiquid AI follow-up on Xx.com/liquidai/status/2122346084186132681 · Nov 16, 2026Tracking continues until an independent board lists it.