Ask HN: Feedback on specialized local LLMs/VLMs

We’ve launched causa™, our application that orchestrates and runs LLMs fully offline across Apple devices (VLM support coming soon). We’re collecting feedback on which specialized or fine-tuned models the community would find most valuable for on-device inference. We already support the main general-purpose families (GPT-OSS, Llama, Mistral, etc.) and are now focusing on domain-specific models fine-tuned for targeted tasks.

Examples: • Mellum series by JetBrains — optimized for software engineering • gpt-oss-safeguard — tailored for policy reasoning and AI safety

If you know of other high-quality specialized models (preferably with open weights) that could benefit from mobile deployment, we’d love your input.