deploy thoxfast 058d0e08b664

Give fast model full CPU quota

Restore Q4_K_M after fidelity failure

Use lower-latency interactive Q4_0

Restore two-context interactive topology

Keep interactive streams live during inference

Bound interactive inference backpressure

deploy latency-bounded interactive pool

THOX Rust Coder: Qwen3-Coder-REAP-25B-A3B-Rust GGUF via conda-forge llama.cpp