Unverified50% confidenceBenchmarkExact time
Across five draft-target pairs built from Llama-3.1-8B-Instruct with Medusa heads, Meta's MTP models, and Qwen3-4B-Inst-2507-MTP, AMTP achieves up to 4.1x speedup over standard decoding.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/21/2026
First Seen
Valid until: 11/4/2026
Sources
Related Entities
Related Claims
Unverified混元3配备约38亿参数的MTP模块，用于推理加速68% similarUnverifiedThrough MTP (Multi-Token Prediction) technology, the inference speed of Qwen3.8-Flash-Next and GLM-5.3-Flash is improved by up to 2x.67% similarUnverifiedAdaptiveSpec supports three mainstream target models: DeepSeek-R1-Distill-Llama-8B, Llama-3.1-8B-Instruct, and Qwen3-8B.66% similarUnverifiedRadixArk/Qwen3.8-Flash-Next-NVFP4 mentioned in the LMSYS blog is an SGLang checkpoint that uses post-training quantization, quantizing only the routed experts, calibrated with 128 CNN/DailyMail articles truncated to 512 tokens, validated on GB300 and B300, and reported to reach 540 tok/s at batch size 1 with B200 TP4 plus MTP.66% similarUnverifiedThe new version combines D-Flash2, MTP, and N-gram speculative decoding, and the combination of all three produces a roughly twofold decoding acceleration effect.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1078993API
curl https://kongchang.com/api/v1/knowledge/claims/1078993MCP
get_claim(id=1078993)