Android
60 updates on Android.
Google lists first Gemini Nano 4 phones, requires Nano 3 for Gemini Intelligence
ML Kit GenAI documentation now names nano-v4 devices and sets Nano v3 or greater, plus 12 GB of RAM, as the requirement for Gemini Intelligence.
Graphcore runs Llama 3.2 11B Vision on a Pixel 8a with 2.7-bit quantisation
Graphcore Research fits Llama-3.2-11B-Vision-Instruct into 3.6 GB with a 2.7-bit format and runs it on a Pixel 8a at 3.8 tok/s.
Ornith releases Ornith-1.5, a 9B model with a mobile build for iPhone and Android
Ornith-1.5 comes in 9B, 35B and 397B sizes, and the 9B model scores 70.6 on SWE-bench Verified and has a mobile build for iPhone and Android.
Online-SDFT reports 70.28% routing accuracy and updates a LoRA adapter on the phone
I-Ju Lin and Zhang-Wei Hong measure 70.28% accuracy against 52.78% for the best baseline, and run the LoRA update itself on an Android phone.
RikkaHub Agent test: Android phone agent compiles whisper.cpp on its own
XDA runs RikkaHub Agent on an Oppo Find N5 against a self-hosted Qwen 3.6 27B. The agent installed dependencies and built whisper.cpp in Termux in seven minutes.
Gemini Nano 4 ships on Samsung foldables with ML Kit Prompt API access
Google says Samsung’s new foldables carry Gemini Nano 4 with support for more than 140 languages, reachable from apps through ML Kit’s Prompt API.
FBLayout fine-tunes transformers on phone GPUs 2.2 to 5.7 times faster
A MobiSys 2026 paper fine-tunes seven transformer models on phone GPUs 2.2 to 5.7 times faster than MNN, TFLite and TVM, with 4.2 times fewer cache misses.
NPUs lead LLM prefill, CPUs lead decoding on Snapdragon phones
A measurement study of five mobile LLM frameworks finds NPUs excel at prefill, CPUs at decoding, and framework gaps of up to 10x on NPUs.
Google adds multi-token prediction to Gemini Nano v3 on Pixel 9 and 10
Google added a prediction head to frozen Gemini Nano v3 on Pixel 9 and 10, saving 130 MB per instance over a standalone drafter.
MLPerf Mobile v6.0 adds Llama tests in 1B, 3B and 8B sizes
MLCommons added Llama tests in 1B, 3B and 8B sizes to its mobile benchmark app, reporting token throughput next to the existing vision and image tests.
llada.cpp runs a diffusion LLM on a Snapdragon NPU up to 42 times faster
Tsinghua and Beihang researchers report LLaDA-8B generating 128 tokens 17 to 42 times faster on a Hexagon NPU than on the phone CPU.
CAPED redacts phone screenshots before a cloud GUI agent sees them
Researchers at CUHK, Southeast University and Shandong University report a phone-side screenshot filter that cuts seeded leakage from 0.766 to 0.268.