Microsoft and OpenAI further accelerate their models for work

Microsoft unveiled seven new AI models at its Ignite conference, while OpenAI introduced two variants of its flagship model. The announcements mainly target professional uses amid significant industry turmoil.

Microsoft presented seven new AI models at its Ignite conference on Tuesday, while OpenAI launched two variants of its flagship model, GPT-6.2 Astra and Sol. Both companies primarily target professional uses, from coding to transcription, as the sector remains affected by financial, legal, and stock market tensions. At this stage, the battle is less about rhetoric and more about cost, speed, and deployability in enterprises.

Des modèles pensés pour le bureau, pas seulement pour les démonstrations

At Microsoft, the new models have very functional names: MAI Cyber 2 Flash for cybersecurity, MAI Code 2 for development, MAI Image 2.7 for image generation, and Dragon Copilot Transcribe for transcription. The publisher aims to make its models more attractive to businesses, with a central argument: they would be less costly to run at scale than competing models from OpenAI and Anthropic. At the same time, Microsoft highlights a heavyweight licensing program that allows clients to tailor the models to their needs. In short, AI should no longer just respond; it must also integrate into business processes without blowing up the bill.

OpenAI follows a different but complementary logic. GPT-6.2 Astra is presented as the most powerful, most expensive, and most computationally demanding version, while Sol targets “everyday” use, meaning daily tasks in ChatGPT and Codex. According to OpenAI, both variants outperform their sibling models in its benchmarks. They have been available to paying subscribers in ChatGPT and Codex worldwide since this week.

Le marché avance, malgré le bruit de fond

These releases come as the AI industry goes through a turbulent phase. The source text mentions outages, cheating scandals, upcoming IPOs, and turmoil around Nvidia and Anthropic. Microsoft and OpenAI seem determined to stay the course, as if the best response to chaos is still to release a new model. It’s not very glamorous, but it aligns with the current logic: show usable products rather than promise magical tomorrows.

From the investors’ side, the contrast is clear. The source mentions a 14% drop in Nvidia’s stock during its first week of turbulence, with a loss of 1 trillion dollars in valuation [to be verified]. Meanwhile, Anthropic is reportedly preparing a potentially historic IPO, with an opening price envisaged at 78 dollars per share and a valuation of 1.4 trillion dollars [to be verified]. These figures mainly say one thing: AI remains a massive betting ground, even when the scenery cracks.

Key points

  • Microsoft presented seven new models at Ignite on Tuesday.
  • OpenAI launched GPT-6.2 Astra and Sol for ChatGPT and Codex.
  • MAI Code 2 and MAI Image 2.7 target work and creation.
  • OpenAI reserves its variants for paying subscribers worldwide.
  • Anthropic is reportedly aiming for an IPO on Thursday, according to the source.

By the numbers

  • 7 models — announced by Microsoft at Ignite on Tuesday.
  • 2 variants — GPT-6.2 Astra and Sol, launched by OpenAI.
  • 14% — reported drop in Nvidia’s stock during its turbulent week [to be verified].
  • 78 dollars — envisaged opening price for Anthropic [to be verified].
  • 1.4 trillion dollars — mentioned valuation for Anthropic [to be verified].

To read