Qwen3.5-9B

9B对话由 Qwen 发布 发起讨论

0

量化版本

67

实测记录

1

讨论数

量化版本

社区与官方发布的量化版本;「关联实测」只统计精确关联到该发布的实测记录。

还没有已收录的量化版本。

实测数据

共 50 组配置

硬件框架量化生成速度运行显存实测次数证据
NVIDIA RTX 5090llama.cppQ2_K277.3 tok/s12.5 GB1L2 跨框架验证
NVIDIA RTX 5090OllamaQ5_K_M166 tok/s19.68 GB1L2 跨框架验证
AMD Ryzen AI Max+ 395vLLMIQ3_XS115.9 tok/s13.82 GB1L2 跨框架验证
AMD Ryzen AI Max+ 395vLLMQ5_K_M83.9 tok/s19.68 GB1L2 跨框架验证
AMD Ryzen AI Max+ 395OllamaQ4_K_S85.5 tok/s16.84 GB1L2 跨框架验证
Apple M5 MaxvLLMIQ3_S222.5 tok/s14.57 GB1L2 跨框架验证
Apple M5 MaxvLLMQ5_K_S163.1 tok/s19.11 GB2L2 跨框架验证
Apple M5 MaxStrataQ3_K_M284.3 tok/s14.95 GB1L2 跨框架验证
NVIDIA DGX Sparkllama.cppQ6_K110.1 tok/s21 GB1L2 跨框架验证
NVIDIA DGX SparkStrataQ3_K_M209.3 tok/s14.95 GB1L2 跨框架验证
NVIDIA Jetson AGX Orin 64GBOllamaQ3_K_S55.4 tok/s14.38 GB1L2 跨框架验证
NVIDIA Jetson AGX Orin 64GBStrataQ4_K_M72 tok/s17.6 GB1L2 跨框架验证
AMD Strix Halo (Radeon 8060S)llama.cppIQ1_S116.2 tok/s10.04 GB1L2 跨框架验证
AMD Strix Halo (Radeon 8060S)llama.cppQ5_K_S60.5 tok/s19.11 GB1L2 跨框架验证
AMD Strix Halo (Radeon 8060S)StrataQ4_K_S105.7 tok/s16.84 GB1L2 跨框架验证
AMD Strix Halo (Radeon 8060S)StrataQ6_K73.1 tok/s21 GB1L2 跨框架验证
2x AMD Strix HaloOllamaFP1658.4 tok/s37.06 GB1L2 跨框架验证
2x AMD Strix HaloStrataIQ3_XS234.6 tok/s13.82 GB1L2 跨框架验证
AMD Strix Point (Radeon 890M)vLLMIQ3_S65.9 tok/s14.57 GB1L2 跨框架验证
AMD Strix Point (Radeon 890M)OllamaIQ3_XS62.9 tok/s13.82 GB1L2 跨框架验证
AMD Strix Point (Radeon 890M)OllamaIQ4_XS46.8 tok/s15.2 GB1L2 跨框架验证
Intel Arc A770 16GBOllamaIQ3_XS56.6 tok/s13.82 GB1L2 跨框架验证
Intel Arc 140VvLLMQ3_K_L54.6 tok/s15.2 GB1L2 跨框架验证
Intel Arc 140VOllamaIQ2_XS61.1 tok/s11.17 GB1L2 跨框架验证
Apple M1llama.cppQ3_K_M51.9 tok/s14.95 GB1L2 跨框架验证
Apple M1StrataIQ1_S96.9 tok/s10.04 GB1L2 跨框架验证
Apple M1 Prollama.cppIQ3_XS75.7 tok/s13.82 GB1L2 跨框架验证
Apple M1 ProvLLMIQ2_M84.2 tok/s11.93 GB1L2 跨框架验证
Apple M1 ProvLLMQ3_K_S79.3 tok/s14.38 GB1L2 跨框架验证
Apple M1 Maxllama.cppQ5_K_M65.4 tok/s19.68 GB1L2 跨框架验证
Apple M1 MaxStrataFP1650.1 tok/s37.06 GB1L2 跨框架验证
Apple M1 MaxStrataQ5_K_M86.9 tok/s19.68 GB1L2 跨框架验证
Apple M2 ProOllamaQ3_K_L72.5 tok/s15.71 GB1L2 跨框架验证
Apple M2 ProOllamaQ5_K_M53 tok/s19.68 GB1L2 跨框架验证
Apple M2 ProStrataQ3_K_M98.4 tok/s14.95 GB1L2 跨框架验证
Apple M2 ProStrataQ5_K_S84.6 tok/s19.11 GB1L2 跨框架验证
Apple M2 Ultrallama.cppQ3_K_S108.4 tok/s14.38 GB1L2 跨框架验证
Apple M2 UltravLLMFP1654.5 tok/s37.06 GB1L2 跨框架验证
Apple M2 UltravLLMQ3_K_L115.9 tok/s15.71 GB1L2 跨框架验证
Qualcomm Snapdragon X EliteStrataQ4_K_S60 tok/s15.2 GB1L2 跨框架验证
NVIDIA RTX 4060 Ti 16GBOllamaIQ3_S73.1 tok/s14.57 GB1L2 跨框架验证
AMD Radeon RX 9070 XTvLLMIQ1_S133.2 tok/s10.04 GB1L2 跨框架验证
AMD Radeon RX 9070 XTvLLMQ3_K_S113.7 tok/s14.38 GB1L2 跨框架验证
AMD Radeon RX 9070 XTOllamaIQ2_XS124.5 tok/s11.17 GB1L2 跨框架验证
AMD Radeon RX 9070OllamaQ2_K96.6 tok/s12.5 GB1L2 跨框架验证
AMD Radeon RX 9070OllamaQ4_K_S80.6 tok/s15.2 GB1L2 跨框架验证
AMD Radeon RX 9070StrataQ4_K_S115.5 tok/s15.2 GB1L2 跨框架验证
AMD Ryzen AI Max+ 395llama.cppIQ3_XS118.3 tok/s13.82 GB1L2 跨框架验证
Apple M4 ProStrataIQ2_M177.4 tok/s11.93 GB1L2 跨框架验证
Apple M5 ProvLLMQ3_K_L145.9 tok/s15.71 GB1L2 跨框架验证

其中 67 条实测未关联到具体量化版本,完整数据见搜索页。

社区讨论

查看全部讨论