Production inference server for open Jev-like decision models (Jev/SystemOne /v1/systemone). Clef-Flash on vLLM: 45 ms p50, 2x throughput, verified against the reference.
cloudflare decision-model inference-server multimodal clef systemone vllm llm-inference qwen3 jev clef-flash
-
Updated
Oct 2, 2026 - Python