Mistral bets on open weights with Large 4
Mistral released the weights of its multimodal Large 4 model on 29 October. In Heise’s KI-Update podcast, Mistral science lead Pierre Stock describes how the French company is using the model to compete in coding and cybersecurity.
Revolut cuts AI model usage costs by 90%
British fintech company Revolut has cut its AI-assisted coding costs by around 90% by switching from commercial services such as Anthropic’s Claude to self-hosted AI models. Initially using GLM-5.1 from Z.ai, the company now runs GLM-5.3 on 56 Nvidia B200 GPUs.
Google introduces Gemini 4 Carbon
Just four weeks after Argon, Google is replacing its flagship model with Gemini 4 Carbon. The new model is expected to deliver significant improvements in coding, autonomous tasks, and cybersecurity.
Anthropic releases Claude Opus 5.5
Anthropic’s Claude Opus 5.5 offers lower pricing, faster results, and strong performance for coding and other demanding professional tasks. Input tokens cost $4 per million, while output tokens cost $20.
Anthropic expands Claude Code
Anthropic has revamped the Projects feature in Claude Code, turning it into a system that can coordinate multiple AI agents. Users provide a goal, and Claude breaks it down into individual tasks and assigns them to parallel threads.
AI Coding needs trust, not speed
The rapid progress in AI-generated code brings a new challenge: trust over speed. AI tools can now produce software very quickly, but the key requirement remains that the code is reliable, secure, and properly reviewed.
SAP and AI coding
SAP CEO Christian Klein has highlighted how AI is transforming software development through so-called “vibe coding.” In this approach, developers describe their requirements in natural language, and AI automatically generates applications in response.
Mistral introduces agentic work interface
Mistral has rebranded its chatbot “Le Chat” as “Vibe” and launched a new agent-based interface. This goes beyond simple conversation and can carry out tasks autonomously. The platform is divided into Chat, Work, and Code sections.
OpenAI’s ultra-fast Coding Model
OpenAI has introduced GPT-5.3-Codex-Spark, a new AI model specifically optimized for programming. Unlike previous versions, it doesn’t run on Nvidia GPUs but on wafer-scale chips from Cerebras Systems, meaning a notable strategic shift.
Anthropic unveils Claude Sonnet 4.6 upgrade
Anthropic has introduced Claude Sonnet 4.6 as the new default AI model for both free and paid users. The update brings significant improvements in logical reasoning, coding, and general computer tasks. All without extra cost.