Source: vLLM GitHub README
Project README for vLLM: origin at UC Berkeley's Sky Computing Lab, feature list, hardware coverage, license.
Contents
- What: open-source project README (read 2026-11-14)
- Author: vLLM project
- Date: 2026-11-14 (accessed)
- Original: vllm-project/vllm
Summary
Describes vLLM as a high-throughput, memory-efficient inference and serving library with an OpenAI-compatible server, started at UC Berkeley’s Sky Computing Lab and now community-maintained.
Key claims
- Origin: Sky Computing Lab, UC Berkeley; over 2,000 contributors; Apache-2.0.
- Features: PagedAttention, continuous batching, chunked prefill, prefix caching, CUDA/HIP graphs and torch.compile, quantization (FP8, INT4/INT8, GPTQ/AWQ, GGUF), speculative decoding, disaggregated prefill/decode/encode, tensor/pipeline/data/expert/context parallelism, structured outputs, tool calling, multi-LoRA, Anthropic Messages API and gRPC.
- 200+ Hugging Face model architectures.
- Hardware: NVIDIA, AMD and Intel GPUs; x86, ARM and PowerPC CPUs; plugins for TPU, Gaudi, Spyre, Ascend and others.