Meta has launched Muse Spark, a new model for Meta AI that adds multimodal reasoning, tool use, image understanding, and parallel task handling.
Anthropic says Mythos Preview can autonomously find and exploit serious software vulnerabilities and showed additional safety concerns in testing.
Google has launched Gemma 4, a new family of open-weight AI models for local use, and replaced its custom Gemma license with Apache 2.0. The release adds four model sizes, stronger multimodal capabilities, support for more than 140 languages, and broader deployment options from phones to workstation GPUs.
Microsoft unveiled MAI AI Models for transcription, voice generation, and image creation as it pushed further toward in-house AI self-sufficiency.
Google has launched Ask Maps in Google Maps while expanding Gemini across Workspace apps with new tools for writing, spreadsheets, slides, and file search.
The Google Workspace CLI allows developers and AI agents to automate tasks across Gmail, Docs, Sheets, Drive, and other Workspace services.
OpenAI updated ChatGPT with GPT-5.3 Instant, then released GPT-5.4 Thinking days later with stronger reasoning, deep web research, computer use, and benchmark gains.
Sarvam opened Indus in limited beta, powered by its 105B sovereign model, and outlined new products across document intelligence, speech, on-device AI, dubbing, and agent infrastructure.
Gemini 3.1 Pro enters preview in February 2026 with adjustable thinking modes, a verified 77.1% ARC-AGI-2 score, and expanded rollout across Gemini and developer tools.
