Hello, I want to make another project with my dual Hailo 10H setup, now with audio pipeline with VAD/ASR/LLM/TTS. Has anyone successfully ran more modern model than Qwen2? Either small Gemma 4 or even Qwen 3.5? 0.8B or 2B variants should run pretty fast if quantized no? Are there some ops/something that make then unable to run on 10H? Thanks!
1 Like
Are you the only person on earth that has an m.2 10H?
Seems like it, but I believe you can get it now from Mouser, Farnell, EBV…
NewEgg has them now.
Hopefully Gemma 4 and Qwen 3.8 will be part of Hailo Model Zoo GenAI v5.4