Unverified50% confidenceBenchmarkExact time
AMD官方基准宣称Halo在GLM Flash 30B的Token生成快40%、Qwen 3.5 122B快12%、GPT-OSS 120B快7%，图像生成比Apple M4 Pro快3到7倍
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/12/2026
First Seen
Valid until: 10/10/2026
Sources
AMD Ryzen AI Halo实测：x86架构能否撼动英伟达DGX Spark？
bilibili量子菠萝_7/7/2026
Related Claims
Unverified在AMD 7900XT显卡上从Ollama换用LM Studio并重新调参后，token生成速度从每秒5提升到3669% similarVerifiedDeepSeek V4 Flash拥有284B总参数，每次激活13B，显存需求约150G63% similarUnverifiedSTM32F103C8T6是意法半导体基于ARM Cortex-M3内核的32位微控制器，主频最高72MHz，拥有64KB Flash和20KB SRAM63% similarUnverifiedAMD MI355X 每 GPU 吞吐量比 NVIDIA B200 SGLang 高 1.25 倍63% similarUnverifiedDeepSeek V4 Flash开启MTP后代码生成速度从31.15 Token/s提升到37.3 Token/s，提升幅度约20%62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/488918API
curl https://kongchang.com/api/v1/knowledge/claims/488918MCP
get_claim(id=488918)