Decision models beat LLMs at structured output: Liquid AI d1 runs 2.7-3.6x faster than LFM3
mervenoyann · x · 2026-11-18
mervenoyann benchmarked Liquid AI's decision model d1 against LFM3 for JSON structured output (warmup excluded), finding a multiplier-level speed edge rooted in architecture: LLMs need prefill + decode to emit JSON, while decision models do a single forward pass.
Results:
- Short input, 1 field: LFM3 298 ms vs d1 83 ms (3.6x faster)
- Short input, 3 fields: LFM3 546 ms vs d1 201 ms (2.7x faster)
For high-frequency structured classification/routing workloads, purpose-built decision models may be a much faster alternative to general-purpose LLMs.
Related event: Liquid AI's Decision Model Beats LLMs at JSON Generation by 2.7-3.6x(2 posts)→
More from Models
- GPT-6.2 Sol tops browser-use leaderboard at 82%, 35% cheaper and 2x faster than predecessor — charliermarsh · 2026-11-18
- Why Anthropic matters: competition, pricing and open source — dh7net · 2026-11-18
- Dev's Deep Dive: GPT-6.2 Astra Improved but Still 'Directionally Very Bad' at Debugging — kimmonismus · 2026-11-18
- OpenAI Charged for a ChatGPT Business Premium Seat for a Month That Can't Be Used — physmathastro · 2026-11-18
- Hunch: Astra 6.2's bad reviews stem from OpenAI dogfooding with unlimited internal swarms — daniel_mac8 · 2026-11-18
- Google rolls out Antigravity Orchestrator: dynamic in-task model switching, 24% less quota — rseroter · 2026-11-18