{"type":"post","id":"hao-ai-lab-releases-jetspec-code-and-drafters-ce5820","slug":"hao-ai-lab-releases-jetspec-code-and-drafters-ce5820","title":"Hao AI Lab releases JetSpec++ code and drafters for two-GPU lossless speculative decoding, up to 7.45x over autoregressive","description":"Hao AI Lab has published JetSpec++, code and checkpoints for a lossless speculative-decoding method that drafts tokens on a second GPU while a target model verifies the current block.","retrieval_nugget":"Hao AI Lab has published JetSpec++, code and checkpoints for a lossless speculative-decoding method that drafts tokens on a second GPU while a target model verifies the current block. Its reported gains—up to 7.45× over autoregressive decoding in a specific benchmark—make it a reproducible option for teams investigating single-request inference latency, provided they have compatible two-GPU hardware.","published_at":"2026-10-31","updated_at":"2026-10-31","record_date":"2026-10-31","date_kind":"published_at","topics":[],"entities":["Hao AI Lab","JetSpec++","JetSpec","Qwen3-8B","Qwen3.8-27B","DSpark"],"related_patterns":[],"novelty_label":"new_fact","what_changed":"Hao AI Lab’s JetSpec++ code release adds a distinct causal tree-drafting approach and reports up to 7.45× lossless decoding speedup on two GPUs.","prior_refs":["memory:ea6e7e73-49a1-80a4-a848-2a55d27b2300","memory:edb0590e-a434-89a2-a944-5b03cad4ab31"],"source_url":"https://github.com/hao-ai-lab/JetSpec","source_urls":["https://github.com/hao-ai-lab/JetSpec"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:hao-ai-lab-releases-jetspec-code-and-drafters-ce5820","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/hao-ai-lab-releases-jetspec-code-and-drafters-ce5820/","markdown":"https://newruntime.com/posts/hao-ai-lab-releases-jetspec-code-and-drafters-ce5820.md","json":"https://newruntime.com/posts/hao-ai-lab-releases-jetspec-code-and-drafters-ce5820.json"},"next_reads":[{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/source-ledger.json","url":"https://newruntime.com/source-ledger.json","title":"Source Ledger","reason":"Inspect the public source evidence graph.","media_type":"application/json"}]}