Decision model face-off: Liquid AI d1.2-ultra matches rivals at half the cost
JosephJacks_ · x · 2026-11-19
Blogger 0xironyAditya benchmarked Liquid AI's new d1.2-ultra against jev 1.14, gpt-6 luna and clef on two real tasks via OpenRouter, 200 picture decisions each: grading storm/fire damage on before/after satellite shots and sorting Serengeti camera-trap bursts.
- Exact damage grade: 47% / 46% / 42% / 41%; send-team-first: 70% / 64% / 68% / 69%; animal-or-empty: 99% / 96% / 96% / 98%; species: 62% / 63% / 62% / 65%
- Cost for all 200 decisions: $0.018 / $0.034 / $0.034 / $0.083 — d1.2-ultra is half the price of jev and luna and 4.7x cheaper than clef
- Catch: d1.2-ultra was the slowest of the four
Takeaway: near-tie accuracy at the top, but d1.2-ultra gets there far cheaper. These models can't replace humans yet, but can pre-sort piles and double-check calls.
More from Models
- Gemini Hydrogen High Reportedly Beats Claude Opus 5.5 Extra Tonight — RylanSchaeffer · 2026-11-19
- Leaked Gemini 4 Neon benchmarks show sweeps across 17 evals vs Claude, GPT — nevrekaraishwa2 · 2026-11-19
- Open DocFieldBench: Gemini 3.8 Flash scores 94.6 on 260-document extraction test — refractr · 2026-11-19
- Kimi K3.1 Pro beats K3 on 14 of 17 benchmarks; pairing with Opus 5.5 halves spend — Kimi_Moonshot · 2026-11-19
- Simular box hits OSWorld v2 SOTA at 80.7% success, $1.79 per task — 8x cheaper than August — xwang_lk · 2026-11-19
- Peking University Releases Foundation Model for Measuring and Predicting Human Behaviour — PekingUniversity · 2026-11-19