OpenAI is testing another new way to expose the complicated processes at work inside large language models. Researchers at the company can make an LLM produce what they call a confession, in which the model explains how it carried out a task and (most of the time) owns up to any bad behavior.

OpenAI is acquiring Neptune, a startup that helps companies track AI model training. The deal value was not disclosed, but reports suggest it is less than $400 million in stock. OpenAI already uses Neptune's tools for its own large language models.

That mainstream deployment shaped user expectations in a way that later transitions struggled to accommodate. In August 2025, when OpenAI initially replaced GPT-4o with its much anticipated then-new model family GPT-5 as ChatGPT’s default and pushed 4o into a “legacy” toggle, the reaction was unusually strong.

OpenAI, the AI research lab, has agreed to acquire Neptune, a startup providing tools for tracking AI model training, enhancing its model training capabilities.

So, bottom line, if OpenAI can substantially reduce the cost of API calls and still deliver AI value, as it seems to have done with GPT-5.1, there's a much better chance it can make the case for including GPT-5.1 in developers' products.

GPT-5.1 Instant is able to leverage newfound “adaptive reasoning” capabilities to determine when it needs to think more carefully about its responses, such as when someone enters a more complex prompt. Improvements have also been made to its instruction following skills, so it will directly address user’s queries with more consistency than before.

OpenAI, valued at $500 billion, is acquiring AI startup Neptune to bolster its model training capabilities and improve language model debugging.

OpenAI characterizes GPT-5.1-Codex-Max as the company’s first coding model explicitly trained to operate across multiple context windows through a technique called compaction. The model can maintain coherent reasoning across millions of tokens, enabling tasks that traditionally exceeded AI systems’ memory limits.

A central focus of the partnership is the development of agentic AI solutions powered by frontier GPT large language models, targeting enterprise clients in sectors such as banking, financial services and insurance (BFSI),

The San Francisco-based company is said to be exploring a plan to lease at least 500 MW of data centre capacity from HyperVault, TCS's newly established data centre arm

OpenAI makes revenue from premium subscriptions to ChatGPT but most users get the free version. The company introduced its own web browser, Atlas, in October, an attempt to compete with Google's Chrome as more internet users rely on AI to answer their questions.