OpenAI Is Building Upgraded GPT Terra Model to Slash AI Costs Across the Board
Android Headlines / Tech News / Artificial Intelligence News
Summary
OpenAI, Anthropic, and Google are preparing a new group of AI models that focus on price, advanced thinking, and visual tasks. OpenAI is testing GPT-6.1 Terra, which aims to perform better while using fewer tokens and costing less. Anthropic is developing a model that better understands what users mean, while Google is upgrading its visual tools with Gemini 4 Carbon.
The AI competition is heating up again. OpenAI, Anthropic, and Google are getting ready to release new models. These include GPT-6.1 Terra, Gemini 4 Carbon, and a new system from the company behind Claude that aims to understand what users want. According to a report, all three companies want their new systems to be not only smarter but also cheaper to run and better at specific tasks.
OpenAI prepares GPT-6.1 Terra to cut costs
OpenAI appears ready to bring back its lower-cost Terra model family. AI researcher Lumina found references to GPT-6.1 Terra in Microsoft's code repositories. These showed that testing settings were being changed from older models to the new one. This kind of behind-the-scenes testing often means a launch is getting close.
The Information reports that the new Terra model aims to perform better than GPT-6.1 Sol while costing less. Sol currently costs $2 for every million input tokens and $10 for every million output tokens. A model that is both cheaper and more capable could help OpenAI compete with Anthropic on price.
Anthropic focuses on understanding users and solving problems creatively
Meanwhile, Anthropic is working on a new model that aims to understand what people mean, even when they do not say it directly. Early testers told The Information that the unreleased model, which might be released under the Fable or Opus name, is good at working out what users really want when their instructions are unclear.
Rather than offering general advice or making simple code fixes, the model can find deeper design problems that cause software bugs and suggest more creative solutions. If Anthropic can make its reasoning reliable and help it go beyond repeating common answers, this could be a major improvement for complex software development tasks.
Google upgrades Gemini with Carbon
Google is also keeping up the competition with Gemini 4 Carbon. According to the report, this model will be an upgrade after the recent launch of Gemini 4 Argon. Testing details suggest that Google remains strong at visual tasks. This could make Carbon especially useful for analyzing videos and automatically testing websites.
Reporter Stephanie Palazzolo noted that some of these launches might be delayed until November. Even so, the main point is clear: AI competition is changing quickly. AI companies are racing to offer tools that are smarter, cheaper, and better at understanding what people need for everyday tasks.