Urgent.News

What's breaking now, across thousands of outlets.

AI

OpenAI Reportedly Releasing Far Smarter and Expensive Internal AI Tomorrow

OpenAI could release new AI models as early as Friday, November 6, 2026, including updated Astra and Sol models, according … Read More The post OpenAI Reportedly Releasing Far Smarter and Expensive Internal AI Tomorrow appeared first on ProPakistani .

OpenAI Reportedly Releasing Far Smarter and Expensive Internal AI Tomorrow

OpenAI may unveil advanced AI models as soon as Friday, November 6, 2026, according to a report by The Information. This potential release includes enhanced versions of Astra and Sol models, following the emergence of an undisclosed internal model that has demonstrated remarkable superiority over GPT-6 Astra in engineering tests.

A separate announcement on social media platform X asserts that this undisclosed internal model will be officially launched on Friday, promising significant performance enhancements over both GPT-6 Astra and GPT-6.1 Sol. Although OpenAI has not officially confirmed this claim, benchmark results from November 3 reveal that the internal model significantly outperforms GPT-6 Astra, scoring an impressive 87.5% on BenchCAD 2.0, compared to the latter's 59.4% at its peak reasoning level.

BenchCAD 2.0 assesses AI proficiency in engineering tasks including 3D models, assemblies, and circuit boards. An X user, @deredleritt3r, provided a detailed comparison of the models' performance and estimated pricing for different reasoning settings. The findings indicate that even the lower estimated setting of the internal model achieves approximately 76% on the benchmark tasks, which is notably higher than the maximum score of 59% achieved by Astra's highest setting.

The internal model's cost per task is approximately $11, whereas Astra's costs range from $0.44 for its lowest setting to $11.05 for its maximum setting. The user also noted that the internal model's pricing estimates were derived from OpenAI's benchmark graph and that the reasoning setting labels are estimations. While these figures establish the internal model's superior performance, they do not guarantee its superiority across all types of tasks.

Additionally, although the X post suggests that the upcoming Sol model will benefit from training techniques employed for the internal model, OpenAI has not yet confirmed the specifics of the Sol model's training or released benchmark comparisons between the two models. OpenAI's potential release of new Astra, Sol, and cybersecurity-focused AI models on Friday remains unconfirmed, and the company has not disclosed the new models' official names, pricing, or availability.

Written by urgent.news from ProPakistani's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at propakistani.pk →

More in AI

Six mistakes in introducing AI automation in SMBs

Six mistakes when introducing AI automation in SMBs Over the past few years, I've been seeing the same pattern. A company of twenty people buys a license, someone builds an impressive demo, everyone…

  • Small companies purchase AI licenses without process clarity
  • Most mistakes occur before coding begins, not during implementation
  • Detailed process table with frequency, time, performer, and error handling prevents failures

LLM Apps in Python: A Production Stack Without Framework Bloat

LLM Apps in Python: A Production Stack Without Framework Bloat The first LLM feature I shipped that actually paid for itself was about 200 lines of Python and one SDK.

  • Minimal Python stack for LLM apps with 200 lines of code
  • Core code consists of HTTP call, tool loop, and state management
  • Framework used only for complex requirements like human-in-the-loop pauses

More from Thursday 5 November →