Brief IA
›

OpenAI Launches Astra and Sol, Already Challenged by Testers

AI Tools·Tom Levy·

OpenAI Launches Astra and Sol, Already Challenged by Testers

OpenAI Launches Astra and Sol, Already Challenged by Testers
Key Takeaways
1OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol this week.
2Their access is limited to paid subscribers in ChatGPT Work and Codex.
3According to Thibault Sottiaux, the models have been modified in recent days, although it is uncertain if they have been re-evaluated.
Why it matters : advancements in benchmark testing coexist with criticisms regarding the use of the models, particularly for professional work.

OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol this week, but developers and early users are criticizing these new models. However, a company official indicates that the versions have been modified recently, with no certainty about whether new tests have been conducted since then.

Criticism and Reservations on the Tested Versions

Ethan Mollick, an associate professor and director of the Generative AI Lab at the Wharton School of the University of Pennsylvania, believes on LinkedIn that Astra is less easily guided and utilized than its competitors, often finding it "lazy." On the same platform, Jason J. Fleagle, head of AI at NETSYNC, ranks GPT-6.2 among the worst models OpenAI has offered.

La lettre du soirGratuit · 19 h

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Thibault Sottiaux, head of OpenAI Codex, raises a reservation on X: influencers had received preliminary versions of Astra and Sol some time ago. According to him, the models have undergone modifications and updates in recent days; he assumes that no one has re-evaluated them since. When asked for comment, OpenAI did not provide an immediate response.

Improved Performance and Paid Access

Paid subscribers can access both models via ChatGPT Work or Codex. Astra maintains the specialization of previous models for complex tasks, while Sol is presented as more suited to everyday needs. Both surpass their predecessors in terms of performance, with benchmark tests indicating gains in several areas.

Some users do not see the value in these new offerings. Feedback also reports possible unforeseen difficulties during more professional uses.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.