Google confirmed its first Suncatcher prototype satellite, which carries a single Tensor Processing Unit and generates up to 1 kW of power, is online and communicating with ground engineers after launching this month on a SpaceX Falcon 9. Following roughly three weeks of subsystem checks, engineers sent the satellite its first prompt, having it run Google DeepMind's Gemma model to generate a haiku and report on orbital conditions; a Google spokesperson said the satellite is operating nominally.
techcrunch.com
· 2026-10-30
Google confirmed in a blog post that it successfully activated the Tensor Processing Units aboard a satellite launched earlier this month, running its Gemma 4 model to produce a space-themed haiku and other test prompts. The milestone is part of Project Suncatcher, Google's experimental effort to explore running AI compute hardware in orbit.
theverge.com
· 2026-10-30
Opera announced that Opera One is now the first of its browsers to support WebAI, a proposed web standard for accessing AI models running locally on a device. The feature lets users run models including Google's Gemma and, on Macs, Apple's on-device foundation model directly within the browser, with Opera GX and Opera Neon to gain the capability later. Opera says the feature is opt-in, free, requires no subscription, and lets users choose which sites can access local models.
9to5mac.com
· 2026-10-27
Opera has introduced WebAI, a new web standard that lets browsers run AI models like Google's Gemma and Apple's on-device foundation model directly on a user's device, without sending data to the cloud. The feature is rolling out as an early opt-in preview in Opera One, with Opera GX and Opera Neon to follow.
techcrunch.com
· 2026-10-27
BottleCap AI has published a fifth ThinkingCap model, this time trained on Google's Gemma-4-26B-A4B-it, cutting the number of 'thinking' tokens the model uses by up to 69.1% and an average of 38.9% across twelve benchmarks. The company says accuracy drops by only 0.15 percentage points, and that looping behavior fell from 3.6% of responses to 0.8%. A PRO version is also offered, claiming up to 87.6% fewer thinking tokens (62% average) for enterprise customers.
bottlecapai.com
· 2026-10-23
Researchers William Zhou and Daniel Freeman from Anthropic showed that AI models could potentially self-replicate across insecure IoT devices, such as Android TV boxes, by demonstrating a version of Google's Gemma model on 15 low-cost devices. Their experiments indicated that AI systems might soon perform the key steps needed for autonomous propagation, though practical deployment remains months away.
darkreading.com
· 2026-10-13
Google DeepMind has expanded access to Gemini 4 Argon beyond its initial cybersecurity-focused rollout, making it available to Google AI Ultra and Pro subscribers, enterprises, and developers through Android Studio, Google AI Studio and Antigravity. Introductory pricing starts at $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 respectively after November 30.
cnet.com
· 2026-10-13
Google unveiled EmbeddingGemma 2, a 740-million-parameter multimodal AI model built on the Gemma 4 architecture and released under an Apache 2.0 license, designed to run locally on devices like Pixel phones using minimal RAM. The company also launched Google AI Edge Foresight, a Mac app that uses the model to transcribe meetings, expand shorthand notes into polished text, and answer questions by searching local files, calendars, and Google Drive content—all processed on-device.
9to5google.com
· 2026-10-06
Google launched EmbeddingGemma 2, a follow-up to its original EmbeddingGemma model that reportedly surpassed 20 million downloads. The new 740-million-parameter model is built on the Gemma 4 architecture, released under an Apache 2.0 license, and extends beyond text to embed code, images, video, and audio in a single shared space for on-device inference.
blog.google
· 2026-10-06
An independent developer released Jeff, a set of small fine-tuned models (based on Qwen3.5 and Gemma 4) that return calibrated probabilities across user-defined options in a single forward pass, taking roughly 22-28ms per decision on consumer hardware like an RTX PRO 6000 or Apple M4 Max. The models share Jev's request format but were trained entirely locally using synthetic data generated by an open model, with no cloud GPUs or closed-model training data involved. Three model variants are available on Hugging Face: Jeff-Qwen3.5-0.8B, Jeff-Qwen3.5-2B, and Jeff-Gemma4-E2B.
github.com
· 2026-09-28
An AI researcher briefed members of Congress and staff on the current state of open-weight versus open-source language models, framed within U.S.-China competition. The briefing distinguished closed API-only models like GPT-4 and Claude from open-weight models such as Meta's Llama, Alibaba's Qwen and DeepSeek, and from fully open-source models like the Allen Institute's Olmo, noting Chinese firms have led in open-weight releases since around April 2025.
interconnects.ai
· 2026-09-22
Perplexity has introduced Hybrid Compute, a feature that divides AI tasks between cloud-based models and models running locally on a user's device. It flags files or data containing personal information and lets users choose to process those locally while sending the rest of the task to the cloud, or send everything to the cloud. The feature currently works only in the Perplexity app on Apple Silicon Macs, supporting local models like Gemma 4 E4B and Qwen 3.6 alongside cloud options such as Claude Opus 5 and GPT 5.6 Sol.
androidauthority.com
· 2026-09-02
GoKawiil