Gemma 4 12B brings native audio to a laptop-sized model
Google released weights under Apache 2.0 with a unified multimodal architecture.
Capvise Media / Stories
AI
The systems learning, generating and acting around us.
90 storiesBrowse by month
Opus 4.8 adds effort controls and cheaper fast-mode access
Claude Code also gains dynamic workflows for larger tasks.
Mistral Medium 3.5 moves Vibe coding tasks into remote sessions
The following week brought work and code under the Vibe name.
Gemini 3.5 Flash becomes the default across Google’s assistant
The I/O launch reached consumers, developers and enterprise customers.
Apple previews AI descriptions and on-device subtitles for accessibility
The May announcement scheduled the software changes for later in 2026.
Anthropic releases ten reference agents for finance workflows
Firms can adapt them to internal policies and approval routes.
Mistral Workflows adds pause-and-resume orchestration for enterprise agents
The preview records execution steps in Studio.
DeepSeek V4 preview brings a million-token context to two model tiers
Open weights and hosted API access arrive through separate deployment routes.
GPT-5.5 reaches ChatGPT and Codex before its API release
OpenAI separated the subscription rollout from API availability.
Decoupled DiLoCo lets training clusters recover more independently
The architecture groups compute into asynchronous learner units.
Opus 4.7 keeps its token rates, but changes how inputs are counted
The model upgrade introduced a migration cost that its headline prices do not fully describe.
Gemini 3.1 Flash TTS adds inline direction for generated speech
Audio tags steer pacing and delivery across more than 70 languages.