vLLM Adds Day-0 Support for Inkling Recurrence Model, Up to 2.68x Faster Decoding on GB200
vLLM announced day-0 support for Thinking Machines' recurrent Inkling model, achieving up to 2.68x faster decoding latency on GB200. A follow-up technical thread explained the architecture: highly shared weights with independent KV states across recurrence passes.
2026-11-21 ~ 2026-11-21 · 2 related posts
- vLLM explains looped Inkling internals: shared weights with separate per-pass KV state — vllm_project · 2026-11-21
- vLLM day-0 support for looped Inkling: 2.68x faster decode on GB200 with 78% fewer kernels — vllm_project · 2026-11-21