Pinku launches private offline AI chat

Pinku makes local AI approachable for mainstream Apple users, trading frontier-model capability for privacy, portability, and predictable offline access.

• –Supports more than a dozen compact models, from Qwen3 0.6B to Qwen3 8B
• –Apple Silicon and Metal acceleration should deliver fast, responsive inference on supported devices
• –Free access lowers the barrier to experimenting with on-device LLMs
• –The privacy story is strong, but model downloads, model lists, and Pro purchases still require network access
• –Smaller local models are useful for drafting, rewriting, and brainstorming, but are not substitutes for cloud-scale reasoning