Microsoft AI executive Nicolas Bustamante has warned against writing off Anthropic, even as OpenAI’s latest AI model, GPT-6 Astra, draws attention for its capabilities in coding, reasoning and autonomous capabilities. Bustamante said that while the GPT-6 Astra is an impressive model, Anthropic could still remain a major player in the growing AI market. He also argued that the competition between AI companies should not be viewed as a race where only one or two companies can succeed.
Microsoft Executive Defends Anthropic
In a post on X, Bustamante pushed back against claims that Anthropic is falling behind its competitors. He wrote, “GPT-6 Astra is an amazing model, but writing off Anthropic is a BIG mistake. A lot of people are calling them dead, behind, falling behind etc. It feels like the exact same mistake people made doubting OpenAI a few months ago. Bottom line: this market is not a fixed pie. Inference will 10,000x from today.”
He added, “The whole frontier is moving up, the market is expanding, and multiple companies can win at the same time. And if you’re struggling to pick the best provider, maybe don’t pick one. Go multi-model.”
Bustamante’s comments suggest that improvements in AI models do not necessarily mean one company will dominate the entire industry. Instead, he expects demand for AI inference and advanced models to increase significantly as businesses and consumers use AI for more tasks.
What Is GPT-6 Astra?
OpenAI’s GPT-6 Astra is described as its latest flagship AI architecture, with a focus on more complex tasks rather than simply answering individual prompts. According to OpenAI, Astra was pretrained using more than 100,000 GPUs. OpenAI claims that the architecture can outperform earlier models such as GPT-5.6 on complete tasks, including coding and logical reasoning.
The model is designed for broader, multistep tasks. Instead of only generating an answer to a user’s prompt, Astra is built to coordinate different steps needed to complete a task. This includes working across everyday computer environments such as browsers, spreadsheets, websites and desktop applications.
A fast mode of GPT-6 Astra is available through the API. According to the information provided, it can deliver up to twice the speed of standard processing, but at twice the standard price.
The listed price per 1 million tokens is:
Input tokens: Rs 940
Cached input: Rs 94
Cache writes: Rs 1,150
Output tokens: Rs 4,700
The pricing structure means developers can choose between standard processing and a faster option depending on their requirements.
GPT-6 Astra vs Claude Fable 5.1
The competition is also expanding as Anthropic continues to develop its own AI models. Based on the information provided, GPT-6 Astra and Claude Fable 5.1 have some overlapping capabilities, but their primary areas of focus differ.
GPT-6 Astra is aimed at computer use, reasoning, coding and autonomous capabilities. It is designed to work across browsers, spreadsheets, websites and desktop applications.
Claude Fable 5.1, meanwhile, focuses on coding, knowledge work, research and long-running AI agents. It can also operate browsers and work across multiple applications. Both models are designed for agentic tasks, where AI systems can use tools and complete multiple steps with limited human intervention.
