Flide speeds up Jev-like decision models 9.5x
The gist
The OmniJev team released Flide, an open inference engine claiming up to 9.5x speedup over vLLM and SGLang.
It targets decision models specifically, a niche where general LLM serving frameworks may lag.
How it unfolded
- Oct 29, 2026Flide's GitHub repository describes it as fast inference infrastructure for Jev-style decision models, with exact state forking making it up to 9.4 times faster than vLLM and SGLang.
- Nov 4, 2026A curated list of decision models describes Flide as an open inference engine from the OmniJev team that forks one state into all of its questions in one forward pass.
- Nov 9, 2026A Show HN post titled "Flide: Make Jev-like decision models up to 9.5x faster" was posted to Hacker News by user unikcc, linking to the project's GitHub page; at scraping time it had 2 points and no comments.
Key terms
- inference engine
- The process by which an AI system applies rules to known facts to derive new conclusions.
- vLLM
- A family of neural network architectures based on multi-head attention that processes tokens with positional information.
Source: Wikipedia