Strategy

Google battles Chinese open-weights models with Gemma 4

Google releases Gemma 4, a 31B multimodal open-weights model that runs on consumer 24GB GPUs, in direct competitive escalation against Chinese open-model leaders Moonshot and Alibaba.

Friday, April 3, 2026 12:00 PM UTC2 MIN READSOURCE: The RegisterBY sys://pipeline

Google DeepMind released Gemma 4, a family of open-weights models under Apache 2.0 license optimized for agentic AI and coding. The flagship 31B model supports multimodal inputs (video, audio), 140+ languages, native function calling, and fits on a single 80GB H100 unquantized or a 24GB consumer GPU at 4-bit. The release is a direct competitive response to a wave of strong Chinese open-weights models from Moonshot AI, Alibaba, and Z.AI.

Read original at The Register

Elon Musk had a bad week in court

Musk's testimony in his lawsuit against OpenAI unraveled in court as he contradicted himself and argued with counsel, potentially crippling his case to reclaim control of the nonprofit.

Policy4d ago

Mozilla's Opposition to Chrome's Prompt API

Mozilla formally opposes Chrome's proposed Prompt API in web standards governance, escalating vendor conflict over browser-initiated dialog patterns.