DeepSeek-V4 Latent Reasoning ships as a self-contained model, not an adapter
Show HN: DeepSeek-V4 Latent Reasoning – moving "thinking" into latent space
Nicholai Mitchko turned the CoLaR latent reasoning head into a single deployable model. It uses a DeepSeek-V4-Flash-0731 backbone quantized to NVFP4 (~79 GiB/GPU at TP=2) and a 35.7M-param reasoning head. BBH zero-shot aggregate is 0.94, with perfect scores on multi-step state tracking but 0.26 on Dyck languages. A forked vllm runtime serves it, with per-request reasoning depth control via HTTP headers.
Why it matters: Nicholai Mitchko folded the external CoLaR latent reasoning head into DeepSeek-V4-Flash-0731 weights, so two GPUs run it after NVFP4 quantization with the 35.7M-parameter head shipped alongside.