zml/llmd 20261112.0

Zml/llmd released version 20261112.0, adding support for Tencent's DSpine drafters alongside more speculative decoding options, faster DeepSeek prefill, Apple M1 and M2 support, and expanded sampling controls. DSpine extends DFlash by passing information between adjacent draft positions as it refines its proposals, and is enabled through the same `--speculative-drafter-model` flag. The release expands the speculative decoding options available to developers running the inference stack.

We are pleased to announce the release of zml/llmd 20261112.0 . This week, we are bringing DSpine, more speculative decoding options, faster DeepSeek prefill, Apple M1 and M2 support, and more control over sampling. This release brings support for Tencent’s DSpine drafters: DSpine extends DFlash by passing information between adjacent draft positions as it refines its proposals. You can use it through the same --speculative-drafter-model flag: