Today for AI
HOT RADAR
llmHEAT 11.5°

AI CURATED HOT STORY · 10/16/2026

PrismML Launches Ternary Bonsai 2 for On-Device AI Agents on iPhone

1 reports archived1 independent sourcesupdated 10/16/2026, 13:50:56
Synthesis & Latest Updates
1 Sources Cross-Validated

PrismML released the Ternary Bonsai 2 series (2B/4B/9B), built on Qwen 3.5 with ternary weight encoding to minimize storage. Designed for edge devices like iPhones and PCs, it achieves up to 120 tokens/sec locally on iPhone 18 Pro Max while supporting tool calling and multimodal tasks under an Apache 2.0 license.

LATEST/PrismML released the Ternary Bonsai 2 series (2B/4B/9B), built on Qwen 3.5 with ternary weight encoding to minimize storage. Designed for edge devices like iPhones and PCs, it achieves up to 120 tokens/sec locally on iPhone 18 Pro Max while supporting tool calling and multimodal tasks under an Apache 2.0 license.

HEAT TRENDHourly heat curve

3 fully observed hours
Now
11.5
Peak
12.014:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

TIMELINECoverage timeline

Total 1 reports · Latest first
  1. IT之家 · 智能时代T2·78 pts
    • The Ternary Bonsai 2 series offers 2B, 4B, and 9B variants with storage as low as 0.43GB, optimized for mobile deployment.
    • On iPhone 18 Pro Max, the 2B model achieves 120 tokens/s locally, while the 9B model reaches 34 tokens/s.
    • Compared to full-precision models, the new series delivers up to 5x throughput and 3x energy efficiency on RTX 4090 and M5 Pro hardware.
    • Model weights are open-sourced under Apache 2.0, supporting deployment via llama.cpp and MLX frameworks.