Tag
AI accounting startup Basis completed a 50-tab tax workbook with GPT-6 Astra in half the time GPT-5.6 Sol needed. Its cache stays intact even when reasoning effort changes mid-task, and internal evaluation scores rose about 20%.
Indian voice and chat agent company Ringg handles more than 7 million calls a month with OpenAI models. Routing each task to a different model cut its model costs by about 90%.
OpenAI fixed the cache-discount reuse window at 30 minutes and shipped a hit-rate dashboard alongside a miss-diagnostic tool. Because cache writes cost 1.25 times the input price, hit rate is now directly tied to cost.
OpenAI opened two models trained with methods similar to GPT-6 Astra. API prices came down by half against GPT-5.6 promotional pricing, and on a business automation benchmark the cost per task ran more than ten times below Claude Opus 5.
OpenAI published a case study on the startup V7 on September 21. It covers V7 Go, which turns company files into context that agents can query, and GPT-6 Astra scored 89% accuracy on the hardest graph queries.
On September 16 OpenAI announced a framework that sets when and how it will disclose cases of misalignment in its models, together with six incident reports from the past six months. Any employee can raise a case, each step has a deadline, and cases are sorted into three tracks. Behind it is the company's own judgment that alignment is not sufficiently solved.
The official ChatGPT account said on September 15 that GPT-5.5 will be pulled from every ChatGPT, ChatGPT Work, and Codex plan on October 14. OpenAI's developer account quickly drew a line, saying the model stays available through the API and Codex sessions authenticated with an API key.
AI watchdog The Midas Project published an analysis on September 10 saying the system cards for GPT-5.6 Preview, GPT-5.6, and GPT-6 Astra lack the loss-of-control risk tier that California's law requires. OpenAI said it is confident it is in compliance.
Specific released Real-SWE on September 12, a benchmark that measures AI coding models on private production codebases licensed from real companies. Fable 5.1 led with a 38.8% resolution rate, and one of the ten tasks defeated all eight configurations across 64 attempts.
Nearly 47,000 people from 186 countries submitted more than 8,000 projects, and on August 25 OpenAI announced eight category winners. Two of the four first-place teams were a veterinarian and a cardiologist with no coding background.
As surging Treasury yields threaten Asia's AI stock rally, AMD's next-generation server chip, a GPT-5.6 price cut, and open-source model and infrastructure discussions filled the day.
That's the last story.
