Whisper turbo vs. medium on Arm and CUDA: The Faster Model Still Flipped Between CPU and GPU A developer benchmarked whisper.cpp v1.9.5 across four builds (x86 CPU, Intel iGPU, Arm CPU, and an RTX Spark CUDA GPU) using the same 6.6-second Japanese recording, finding that medium beat large-v3-turbo on both CPUs while turbo was faster or tied on both GPUs. Per-phase timings explained the flip consistently across all four setups, supporting the explanation that the model with the smaller encoder wins where the encoder dominates. The developer also found that building the Arm CPU version with MSVC, which lacked NEON, made inference roughly 6x slower than a clang/NEON build. whisper.cpp v1.9.5, one recording, four builds x86 CPU, Intel iGPU, Arm CPU, RTX Spark CUDA : medium was faster on both CPUs, turbo was faster or tied on both GPUs, and the per-phase timings had the same shape on all four. This is an English adaptation of an article I originally published in Japanese on Zenn: whisper の turbo と medium、Arm の CPU と CUDA でも速い順は入れ替わった https://zenn.dev/maruhana/articles/smartwatch-whisper-arm-cuda-four-builds . The measurements and conclusions are mine; I used an AI tool to help translate and restructure the text, and I checked the numbers against the original. Short version: on an x86 CPU and an Arm CPU, medium was faster than large-v3-turbo; on two different GPUs, turbo was faster or tied in total , and the per-phase timings explained the flip the same way on all four. A 67-second video summary narration is synthetic, English captions available : system info shows it. large-v3-turbo fp16 , medium , and turbo's q5 0 quantization. -t 8 vs. -t 20 on the Arm CPU. This continues an earlier article in Japanese : whisper の turbo と medium、速いほうが CPU と内蔵 GPU で入れ替わった https://zenn.dev/maruhana/articles/smartwatch-whisper-turbo-cpu-slower . That one covers the method same recording, same arguments, each condition run 5 times in a row with the first run discarded and how to read the phase timings. There, on one PC, medium was faster on the CPU and turbo was faster on the integrated GPU. This time I added a second PC with a different CPU architecture and a different GPU family. whisper.cpp at v1.9.5 on both machines, same recording, same arguments. Average of runs 2–5, in ms. | Model | x86 CPU | x86 iGPU | Arm CPU | Arm CUDA | |---|---|---|---|---| | large-v3-turbo fp16 | 13,192 | 2,393 | 13,873 | 1,537 | | medium | 10,370 | 2,408 | 9,698 | 1,794 | | large-v3-turbo q5 0 | 12,021 | 1,371 | 7,108 | 795 | | Faster of turbo and medium | medium 1.27× | tie 15ms apart | medium 1.43× | turbo by 257ms | The previous comparison was within one PC, so I couldn't tell whether the explanation "where the encoder is heavy, the model with the smaller encoder wins" was specific to that PC. The same direction showed up on a CPU with a different instruction set and on a different family of GPU, so the explanation now covers a little more ground. Both machines used the same recording Japanese speech recorded with the watch's microphone, 6.635 s, 16kHz, mono, 16-bit and the same arguments: whisper-cli -m