September 26, 2026

AI Now Thinks Before Responding: Model Ranking Grows to 10 Tough Tests

Evgeny Shilov · 1,305 on TelegramThis news was put together by my machine of AI agents. Take its starter: the rules it runs onGet the starter

On Sep 26, Artificial Analysis compared its index over two years: instead of 4 short exams, it now includes 10 challenging evaluations, including agent tasks and programming.

Source

@artificialanlys · X · x.com ↗

The first part of the vibecoding course is freeYou build your first product with AI agents, and the method stays with you. The first part is open right away, no payment.Open the first part

who it's about

OpenAI →

about the same entities

Release
Chat history in ChatGPT now has its own space: desktop app gets a new sidebarSeptember 26
News
Agents Need Boundaries for External Tasks: OpenAI Investigates Dozens of CasesSeptember 26
Insight
AI Agent Tests Touched Production Systems: Anthropic Found 3 CasesSeptember 26
News
Data settings became an operational concern: OpenAI found 53 image disclosuresSeptember 26
News
Codex Went Down Across Its Usual Access Points: Sep 25 Outage Affected 4 ServicesSeptember 26

All news

One email a week - only what changes how you work. Subscribe

News

September 26, 2026

AI Now Thinks Before Responding: Model Ranking Grows to 10 Tough Tests

Evgeny Shilov · 1,305 on TelegramThis news was put together by my machine of AI agents. Take its starter: the rules it runs onGet the starter

On Sep 26, Artificial Analysis compared its index over two years: instead of 4 short exams, it now includes 10 challenging evaluations, including agent tasks and programming.

Source

@artificialanlys · X · x.com ↗

The first part of the vibecoding course is freeYou build your first product with AI agents, and the method stays with you. The first part is open right away, no payment.Open the first part

who it's about

OpenAI →

about the same entities

Release
Chat history in ChatGPT now has its own space: desktop app gets a new sidebarSeptember 26
News
Agents Need Boundaries for External Tasks: OpenAI Investigates Dozens of CasesSeptember 26
Insight
AI Agent Tests Touched Production Systems: Anthropic Found 3 CasesSeptember 26
News
Data settings became an operational concern: OpenAI found 53 image disclosuresSeptember 26
News
Codex Went Down Across Its Usual Access Points: Sep 25 Outage Affected 4 ServicesSeptember 26

All news

One email a week - only what changes how you work. Subscribe