QLLM Claims to Achieve O(1) Inference
ExtremeKangaroo5437 · reddit · 2026-07-09
The author introduces QLLM, a new architecture proposed by their team that uses neither transformers nor mamba, claiming features like O(1) inference and no need for a KV cache.
The post mentions that the theory was developed first, followed by a trainable practical version. They subsequently collaborated with researchers from UC Berkeley and Indiana University to publish a paper, eventually building a 100M parameter model.
More from Models
- Claim says Kimi was distilled from Fable, sparking a model-attribution jab — cephaloform · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- OpenAI’s Codex + GPT-5.6 Sol hits 99% recall in Project APE verification tests — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Macaron V1 adds LoRA RL on GLM 5.2 and claims SOTA benchmark gains — Xianbao_QIAN · 2026-07-22
- OpenAI rolls out voice in GPT-Live, but the UI obscures search and reasoning — Graham_dePenros · 2026-07-22