- Dotika
- Posts
- OpenAI releases GPT-5.1 update
OpenAI releases GPT-5.1 update
ALSO : Google invests €5.5 billion in Germany
Geoffrey NICHIL
November 13, 2025
News of the day
1. OpenAI's GPT-5.1 update enhances reasoning, personalization, and safety with new Instant and Thinking variants and user-adjustable styles.. → Read more
2. Google commits €5.5 billion to Germany by 2029, expanding AI/cloud infrastructure, creating jobs, and boosting GDP, with a focus on data centers and digital skills. → Read more
3. Luxembourg launches Vibe Coding, a 2-day public training event where participants build a functional app using AI collaboration by Dec 8-9. Limited spots.→ Read more
4. OpenAI's new open-weight models are being adopted by the US military for secure, offline operations, potentially boosting competition and specialized AI development.→ Read more
Our take
Hi Dotikers!
This week, OpenAI is rolling out GPT-5.1, an upgrade that isn’t aiming for a dramatic breakthrough but finally focuses on what actually matters to users: the quality of interaction and control over tone. Two variants are arriving ; Instant for everyday conversation and Thinking for complex requests. The flagship promise is adaptive reasoning, a mechanism that adjusts thinking time based on task difficulty. It’s less spectacular than a raw performance leap, but far more useful in daily use.
The strategic move lies elsewhere. ChatGPT is becoming customizable at the account level, with predefined styles and more granular settings for conciseness, temperature, and more. In practice, your preferences apply across all conversations, including ongoing ones, and the assistant can suggest updates in context. This shifts part of “prompting” toward persistent parameters. Done well, it’s a real productivity boost. Poorly governed, it becomes a compliance and trust liability. Yes, the “Cynical” option exists ; but it shouldn’t replace your Monday-morning colleague.
On the safety front, OpenAI is publishing a system-card addendum with new evaluation milestones, particularly around mental health and emotional dependency. Results remain broadly in line with GPT-5, with improvements in jailbreak robustness on the Instant side and a few targeted regressions on the Thinking side that will require further iterations. Transparency on these metrics is improving, but there is still a need for stronger guarantees regarding preference governance and the use of data tied to personalization.
In the short term, automatic routing between variants and a three-month transition window for older models make the adoption path clearer. The bottom line is simple: GPT-5.1 doesn’t redefine the model ; it makes it more manageable. And in a saturated market, manageability is a weapon.
G.
Meme of the day
Reply