Generative AI
Generative models: reasoning, multimodality, alignment and transformation
Overview
Development of supervised generative systems for reasoning, multimodality, value alignment and cognitive uses.
Impact: Acceleration of selected cognitive tasks, with alignment and governance challenges
Projects
Active
OpenAI — GPT-5.6
OpenAI
Available in ChatGPT, Codex and the OpenAI API in Sol, Terra and Luna variants.
A general-purpose model family available in ChatGPT, Codex and the OpenAI API, with three tiers for different capability, cost and latency trade-offs.
- Configurable reasoning
- Knowledge work
- Programming
- Agentic workflows
Roadmap: No dated public future milestone; next review upon an official announcement.
2026-07-09
Google DeepMind — Gemini 3.7 Flash
Google DeepMind
Model card published on 13 August 2026; distributed through Gemini, Google AI Studio, the Gemini API and Google platforms.
An iteration of the Gemini 3 family based on Gemini 3.6 Flash, with adjustable reasoning configurations to balance quality, cost and latency.
- Configurable reasoning
- Text, image, audio and video input
- Multimodal context
- Distribution through Google APIs and applications
Roadmap: No dated public future milestone; next review upon an official announcement.
2026-08-13
Anthropic — Claude Opus 5
Anthropic
Anthropic made Claude Opus 5 available on all its platforms on 24 July 2026 under the claude-opus-5 API identifier.
A Claude model for complex agentic development and enterprise work, available on Claude platforms and through the API.
- Complex agentic programming
- Analysis and enterprise work
- Long-running multi-step tasks
- Tool use
Roadmap: No dated public future milestone; next review upon an official announcement.
2026-07-24