AI News Roundup: Mistral Large 4, Gemini 4 Carbon, and an Open Speech Model from Nvidia

Mistral opens the weights of a trillion-parameter model, Google ships a cyber-focused Gemini, Nvidia releases a free transcription model, OpenAI rewrites its Model Spec, and AWS posts its fastest growth in years.

Here is the AI news from the last day or so that matters for small businesses using AI agents. Every item links to its source.

Mistral releases Mistral Large 4 as open weights

Mistral put the weights for Mistral Large 4, a 1-trillion-parameter multimodal model nicknamed Le Chonk, on Hugging Face. It accepts text and images and can be self-hosted. (Digest AI)

Google introduces Gemini 4 Carbon

Google's new model is built for long engineering and cybersecurity work and topped the Artificial Analysis Intelligence Index v5 at 59. For now, only selected cyber defenders get access. (The Neuron)

Nvidia ships a free speech-to-text model

Nemotron 3.5 Transcribe is an open-weight model that transcribes 40 languages, translates 31 of them to or from English, and labels up to eight speakers in recordings up to an hour long. (AI Early Edition)

OpenAI updates its Model Spec

OpenAI reorganized the public rules for how its models behave around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. (OpenAI)

Gemini can now create and edit files

Google rolled out document, PDF, spreadsheet and slide creation and editing directly inside Gemini chats for all users. (ModelWool)

AWS grows 39%

Amazon said AWS sales rose 39% to $46.0 billion in the third quarter, its fastest growth in nearly five years. (Digest AI)

What it means for your sales calls

Free, accurate transcription in dozens of languages makes it cheaper to record, review and learn from every sales call, including calls in Spanish. That's the kind of insight Ozkra turns into action: AI handles the first conversation, and a human closer takes the warm leads. Want to see it on your calls? Talk to us.