vllm-qwen3-core-optimization

What problem does it solve?

PR-backed optimization for Qwen3 Core in vLLM, enabling auditors, engineers, and researchers to systematically review and enhance dense and MoE models, embeddings/rerankers, and quantization paths while documenting Eagle3 speculative decoding workflows.

Core Features & Use Cases

• PR-driven performance improvements for Qwen3 Core and Qwen3 MoE runtimes in vLLM.
• Guidance for embedding/reranker integration, GGUF/GPTQ/ModelOpt quantization paths, and Eagle3 speculative decoding.
• Use Case: as part of a release readiness review, run through a PR history and runtime surfaces to validate optimizations before production deployment.

Quick Start

Provide an optimization plan for a specified Qwen3 Core model in the vLLM environment, including a review of PR notes and runtime surfaces.