vLLM
VerifiedHigh-throughput, low-latency LLM serving engine powered by PagedAttention.
vLLM Chronological Timeline
v0.29.1rc0
vLLM v0.29.1rc0 introduces dual-key Gumbel-max watermarking support specifically for speculative decoding workflows. This implementation enables cryptographically verifiable provenance for model outputs generated through speculative execution pipelines.
proto-v0.1.0
The vLLM project has released vllm-proto version 0.1.0. This release introduces the initial implementation of the proto package within the vLLM ecosystem.
v0.29.0
Model Runner V2 (MRV2) is now the default execution engine for all models in vLLM. This release introduces CUDA graph memory profiling for KV cache auto-sizing and batch-sharded sampling to reduce per-step logits memory usage.
v0.29.0rc6
vLLM v0.29.0rc6 introduces a default dense prefix cache configuration for hybrid model architectures. This update addresses issue #55 to ensure consistent memory management across mixed-model deployments.
v0.29.0rc5
<p>[Core] Default prefix_cache_retention_interval to dense for Mamba + E…</p>...
v0.29.0rc4: [Bugfix] Avoid sync in TRT-LLM ragged prefill
<p>Generated-by: Codex <a href="mailto:codex@openai.com">codex@openai.com</a></p> <p>Signed-off-by: Codex <a href="mailto:codex@openai.com">codex@openai.com</a></p>...
Complete vLLM Change Log Index
| Date | Change Title | Type | Impact | Details |
|---|---|---|---|---|
| Sep 12, 2026 | v0.29.1rc0 | feature | 8/10 | View ➔ |
| Sep 11, 2026 | proto-v0.1.0 | product_launch | 5/10 | View ➔ |
| Sep 10, 2026 | v0.29.0 | feature | 9/10 | View ➔ |
| Sep 8, 2026 | v0.29.0rc6 | feature | 7/10 | View ➔ |
| Sep 8, 2026 | v0.29.0rc5 | feature | 8/10 | View ➔ |
| Sep 4, 2026 | v0.29.0rc4: [Bugfix] Avoid sync in TRT-LLM ragged prefill | feature | 8/10 | View ➔ |