Unverified50% confidenceFactExact time
Model Runner V2 gains two capabilities in vLLM v0.28.0: E/P/D separation (Encode/Prefill/Decode disaggregated deployment) and weight offloading.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedvLLM v0.28.0rc2 introduces a new speculative decoding feature called DFlash2, which combines two major technical components: local convolution and a candidate selector (PR #52816).68% similarUnverifiedModel Runner V2 (MRV2) capabilities are extended to non-generative workloads, supporting encoder-only attention, sequence pooling, BGE-M3 pooling, and multimodal execution on CPU.66% similarUnverifiedvLLM v0.27.0 introduces a simplified fault-tolerance framework designed specifically for DP+EP external load balancer deployment scenarios, and provides asynchronous preparation capability for elastic EP scaling.66% similarUnverifiedvLLM v0.28.0 introduces DFlash2 and DSpark confidence-scheduled verification in the speculative decoding domain.63% similarUnverifiedvLLM announced support for Speculative Decoding technology on AMD GPUs.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/912417API
curl https://kongchang.com/api/v1/knowledge/claims/912417MCP
get_claim(id=912417)