🔥 Bonfyre

Model compression that actually runs inference — and 50 C binaries that replace your entire backend. Two flagships. One stack. MIT licensed.

Model compression
0.997Inference cosine
50Backend binaries
20Live apps
~2.1 MBTotal disk
🚀

What Can Bonfyre Do For Me?

See what Bonfyre does, who it's for, and real-world uses for your business — restaurants, law offices, agencies, podcasters, and more. No technical jargon.
Includes:
What it replaces · Industry recipes · Live apps · WordPress · Pipelines · Hardware guide · Use cases
Explore Bonfyre →
🔧

Technical Docs & Benchmarks

FPQ compression deep-dive, FPQ-X operators, KV cache, hardware fits with pricing, architecture, all 50 binaries, install, and reproduction commands.
Includes:
FPQ · FPQ-X · KV cache · Benchmarks · Hardware fits · Architecture · Install · Binaries
Open Technical Docs →