Qwen3-Coder-Next

代码由 Qwen 发布 发起讨论

0

量化版本

62

实测记录

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 5090OllamaIQ2_M299.2 tok/s1.05 GB1L2 跨框架验证
NVIDIA RTX 5090OllamaQ3_K_M238.2 tok/s1.05 GB1L2 跨框架验证
NVIDIA RTX 5090StrataIQ1_S320.2 tok/s1.05 GB1L2 跨框架验证
AMD Ryzen AI Max+ 395llama.cppQ4_K_S90.7 tok/s1.05 GB1L2 跨框架验证
AMD Ryzen AI Max+ 395OllamaIQ2_M108.9 tok/s1.05 GB1L2 跨框架验证
AMD Ryzen AI Max+ 395OllamaQ3_K_S98.6 tok/s1.05 GB1L2 跨框架验证
Apple M5 Maxllama.cppQ5_K_S144.1 tok/s1.05 GB1L2 跨框架验证
Apple M5 MaxStrataQ2_K312.3 tok/s1.05 GB1L2 跨框架验证
NVIDIA DGX SparkOllamaQ3_K_L131.5 tok/s1.05 GB1L2 跨框架验证
NVIDIA DGX SparkOllamaQ4_K_S128.8 tok/s1.05 GB1L2 跨框架验证
NVIDIA DGX SparkStrataIQ2_M261.6 tok/s1.05 GB1L2 跨框架验证
NVIDIA DGX SparkStrataIQ2_XS259.5 tok/s1.05 GB1L2 跨框架验证
NVIDIA Jetson AGX Orin 64GBllama.cppQ4_K_S56.6 tok/s1.05 GB1L2 跨框架验证
NVIDIA Jetson AGX Orin 64GBOllamaQ3_K_M54.9 tok/s1.05 GB1L2 跨框架验证
AMD Strix Halo (Radeon 8060S)llama.cppQ8_046.3 tok/s1.05 GB1L2 跨框架验证
AMD Strix Halo (Radeon 8060S)OllamaQ2_K85.3 tok/s1.05 GB1L2 跨框架验证
2x AMD Strix HalovLLMQ3_K_S149.2 tok/s1.05 GB1L2 跨框架验证
2x AMD Strix HalovLLMQ8_087.7 tok/s1.05 GB1L2 跨框架验证
2x AMD Strix HaloStrataQ3_K_M208.4 tok/s1.05 GB1L2 跨框架验证
AMD Strix Point (Radeon 890M)vLLMIQ1_S79.9 tok/s1.05 GB1L2 跨框架验证
AMD Strix Point (Radeon 890M)vLLMQ8_038.5 tok/s1.05 GB1L2 跨框架验证
AMD Strix Point (Radeon 890M)OllamaQ5_K_M41.6 tok/s1.05 GB1L2 跨框架验证
Intel Arc A770 16GBvLLMIQ3_S59.2 tok/s1.05 GB1L2 跨框架验证
Intel Arc A770 16GBvLLMQ4_K_S49 tok/s1.05 GB1L2 跨框架验证
Intel Arc A770 16GBOllamaIQ2_XS65.5 tok/s1.05 GB1L2 跨框架验证
Intel Arc A770 16GBOllamaIQ3_XS57 tok/s1.05 GB1L2 跨框架验证
Intel Arc 140VvLLMIQ4_XS47.9 tok/s1.05 GB1L2 跨框架验证
Intel Arc 140VOllamaIQ4_XS45.3 tok/s1.05 GB1L2 跨框架验证
Intel Arc 140VOllamaQ4_K_S43.9 tok/s1.05 GB1L2 跨框架验证
Apple M1vLLMIQ4_XS47.9 tok/s1.05 GB1L2 跨框架验证
Apple M1OllamaFP1621.4 tok/s1.05 GB1L2 跨框架验证
Apple M1OllamaQ3_K_L41.5 tok/s1.05 GB1L2 跨框架验证
Apple M1 ProvLLMQ6_K50.3 tok/s1.05 GB1L2 跨框架验证
Apple M1 ProOllamaQ3_K_S67.5 tok/s1.05 GB1L2 跨框架验证
Apple M1 MaxOllamaIQ4_XS76.5 tok/s1.05 GB1L2 跨框架验证
Apple M1 MaxStrataIQ2_M143.3 tok/s1.05 GB1L2 跨框架验证
Apple M2 Prollama.cppQ5_K_M54.5 tok/s1.05 GB1L2 跨框架验证
Apple M2 ProvLLMQ3_K_S89.6 tok/s1.05 GB1L2 跨框架验证
Apple M2 Ultrallama.cppQ4_K_S104.6 tok/s1.05 GB1L2 跨框架验证
Apple M2 UltravLLMIQ1_S161.8 tok/s1.05 GB1L2 跨框架验证
Apple M2 UltravLLMQ5_K_S96.1 tok/s1.05 GB1L2 跨框架验证
Qualcomm Snapdragon X Elitellama.cppQ3_K_L49 tok/s1.05 GB1L2 跨框架验证
Qualcomm Snapdragon X EliteOllamaIQ2_M54.2 tok/s1.05 GB1L2 跨框架验证
Qualcomm Snapdragon 7 Gen 4llama.cppIQ1_S53.9 tok/s1.05 GB1L2 跨框架验证
Qualcomm Snapdragon 7 Gen 4OllamaIQ3_S37.5 tok/s1.05 GB1L2 跨框架验证
Qualcomm Snapdragon 7 Gen 4OllamaQ4_K_M32.7 tok/s1.05 GB1L2 跨框架验证
NVIDIA RTX 3090vLLMQ5_K_M——1L0 自报
NVIDIA RTX 3090vLLMQ4_K_M——1L0 自报
NVIDIA RTX 3090llama.cppQ5_K_M——1L0 自报
NVIDIA RTX 3090llama.cppQ4_K_M——1L0 自报

其中 62 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

还没有讨论——发起第一个话题吧。

发起讨论