What is the "Best LLM for Shenava-Koochik ASR"

Shenava-Koochik, a 114M-parameter Persian streaming ASR model from Reza2kn's Shenava collection, is not the compute bottleneck in Persian speech pipelines; the LLM decision/intent stage is where failures occur, according to an analysis of the model. For Persian intent and decision-making, general multilingual models such as Qwen2.5 and Gemma outperform most small English-first models in Farsi, while PartAI/Dorna-Llama3-8B-Instruct is fine-tuned specifically for Persian instruction-following and reasoning. On Reachy Mini's onboard Pi, an 8B model is likely too heavy for real-time local inference, so smaller Qwen2.5 1.5B/3B models or routing the LLM call to the HF Inference API/Endpoints is recommended while keeping ASR local.

Worth noting Shenava-Koochik itself is a 114M Persian streaming ASR model from Reza2kn’s Shenava collection , so ASR shouldn’t be the compute bottleneck here, the LLM stage is where “disition” decision/intent understanding is failing. For Persian-language intent/decision-making specifically, general multilingual models like Qwen2.5 or Gemma tend to do noticeably better in Farsi than most “small” English-first models, that’s usually the actual fix rather than the ASR choice. If you want something Persian-specialized instead, PartAI/Dorna-Llama3-8B-Instruct is fine-tuned specifically for Persian instruction-following and reasoning. If you’re running this on Reachy Mini’s onboard Pi, 8B is probably too heavy for real-time local inference though, you’d want a smaller Qwen2.5 1.5B/3B or route the LLM call to HF Inference API/Endpoints instead of running it on-device, keeping ASR local and offloading only the decision step.