https://freedom.tech/project/vllm/ Every vLLM release and update tracked by Freedom.Tech. en Freedom.Tech is a Foundation project. https://foundation.xyz Freedom.Tech, a Foundation project (https://foundation.xyz) https://freedom.tech/posts/2026-08-26-vllm-0-28-0/ https://freedom.tech/posts/2026-08-26-vllm-0-28-0/ Wed, 26 Aug 2026 09:46:30 GMT 584 commits from 270 https://freedom.tech/posts/2026-08-11-vllm-0-27-1/ https://freedom.tech/posts/2026-08-11-vllm-0-27-1/ Tue, 11 Aug 2026 10:47:49 GMT VLLM adds support for quantized DSpark Markov heads, enabling efficient inference with compressed attention mechanisms https://freedom.tech/posts/2026-08-10-vllm-0-27-0/ https://freedom.tech/posts/2026-08-10-vllm-0-27-0/ Mon, 10 Aug 2026 21:18:11 GMT 561 commits from 242; Review the changelog before upgrading production inference workloads https://freedom.tech/posts/2026-07-27-vllm-0-26-0/ https://freedom.tech/posts/2026-07-27-vllm-0-26-0/ Mon, 27 Jul 2026 01:06:58 GMT VLLM 0.26.0 adds Inkling model support, DeepSeek-V4 performance optimizations, fp32 generation heads, flexible attention backends, and mature KV offloading with tiered storage https://freedom.tech/posts/2026-07-25-vllm-0-26-0/ https://freedom.tech/posts/2026-07-25-vllm-0-26-0/ Sat, 25 Jul 2026 10:38:53 GMT 411 commits from 212; Check release notes for model and hardware support changes https://freedom.tech/posts/2026-07-14-vllm-0-25-1/ https://freedom.tech/posts/2026-07-14-vllm-0-25-1/ Tue, 14 Jul 2026 08:51:20 GMT VLLM 0.25.1 fixes FFmpeg import blocking model startup and mixed-dtype allreduce corruption in quantized models https://freedom.tech/posts/2026-07-11-vllm-0-25-0/ https://freedom.tech/posts/2026-07-11-vllm-0-25-0/ Sat, 11 Jul 2026 20:06:44 GMT 558 commits, 232; Compatibility and performance sweep https://freedom.tech/posts/2026-06-29-vllm-0-24-0/ https://freedom.tech/posts/2026-06-29-vllm-0-24-0/ Mon, 29 Jun 2026 19:41:59 GMT 571 commits from 256; No breaking changes called out in summary https://freedom.tech/posts/2026-06-15-vllm-0-23-0/ https://freedom.tech/posts/2026-06-15-vllm-0-23-0/ Mon, 15 Jun 2026 05:27:20 GMT DeepSeek-V4 hardening, Model Runner V2 expansion to Llama/Mistral, Rust frontend maturation with streaming and LoRA endpoints, Gemma 4 support, Transformers v5 compatibility, multi-tier KV cache offloading https://freedom.tech/posts/2026-06-12-vllm-0-23-0/ https://freedom.tech/posts/2026-06-12-vllm-0-23-0/ Fri, 12 Jun 2026 23:29:26 GMT Minimax M3 models remain unsupported in this version https://freedom.tech/posts/2026-06-05-vllm-0-22-1/ https://freedom.tech/posts/2026-06-05-vllm-0-22-1/ Fri, 05 Jun 2026 10:10:00 GMT VLLM 0.22.1 adds Mellum v2 support, enables zentorch-accelerated quantized inference on AMD Zen CPUs, and fixes multi-node Ray serving hangs and model initialization regressions https://freedom.tech/posts/2026-05-29-vllm-0-22-0/ https://freedom.tech/posts/2026-05-29-vllm-0-22-0/ Fri, 29 May 2026 10:28:13 GMT VLLM 0.22.0 hardens DeepSeek V4 with fused kernels and CUDA graphs, advances Model Runner V2 toward default, adds experimental Rust frontend, and enables multi-tier KV cache offloading for extended memory management https://freedom.tech/posts/2026-05-15-vllm-0-21-0/ https://freedom.tech/posts/2026-05-15-vllm-0-21-0/ Fri, 15 May 2026 08:44:26 GMT VLLM 0.21.0 adds KV offload + hybrid memory allocator integration, speculative decoding with thinking budgets, TOKENSPEED_MLA for Blackwell, and requires C++20 and Transformers v5 https://freedom.tech/posts/2026-05-10-vllm-0-20-2/ https://freedom.tech/posts/2026-05-10-vllm-0-20-2/ Sun, 10 May 2026 07:37:57 GMT VLLM 0.20.2 fixes DeepSeek V4 sparse attention hangs, KV cache allocation failures, gpt-oss torch.compile compatibility, and Qwen3-VL load issues