LMDeploy v0.15.0 发布：支持 DeepSeek V4 并优化推理性能

AI 摘要

LMDeploy 发布 v0.15.0 版本，新增对 DeepSeek V4 的支持，并引入长上下文和 MTP 前缀缓存命中、推测解码的引导解码、内存分配器与调度器集成等特性。同时优化了 TTFT、流式响应解析、FP8 MoE 等性能，并修复了多项 bug，包括前缀缓存、Triton FP8 通信、多模态工具消息解析等。

正文节选

<!-- Release notes generated using configuration in .github/release.yml at main --> ## What's Changed ### 🚀 Features * Support long-context and MTP prefix-cache hits by @grimoire in https://github.com/InternLM/lmdeploy/pull/4688 * [Feature] Add guided decoding support for speculative decoding by @windreamer in https://github.com/InternLM/lmdeploy/pull/4559 * feat(turbomind): memory allocator, object cache, and scheduler integration by @lzhangzz in https://github.com/InternLM/lmdeploy/pull

发布时间：2026-07-31 21:00

抓取时间：2026-08-31 00:45

来源机构：Shanghai AI Laboratory