Open-Source hunch Delivers Millisecond Local Decisions on EmbeddingGemma 2

Open-source project hunch builds a real-time local semantic decision system on Google's EmbeddingGemma 2, delivering text judgments in 22ms on CPU at under 3 cents per million tokens, earning a reshare from the official Gemma account.

2026-11-05 ~ 2026-11-05 · 3 related posts

• Community project turns EmbeddingGemma 2 into an on-device classifier at ~20ms per text input — googlegemma · 2026-11-05
• hunch Turns EmbeddingGemma 2 Into a Real-Time Local Semantic Decision System: 22ms on CPU, Under 3 Cents Per Million Tokens — aigclink · 2026-11-05
• Open-Source hunch Distills LLM Judgments Onto CPU: Millisecond Decisions Under $0.03/Million Tokens — aigclink · 2026-11-05