The new GPT‑5 model can handle prompts of up to a million tokens, allowing developers to generate far longer and more intricate text outputs than previous versions.
OpenAI has announced the launch of GPT‑5, its most powerful language model yet, featuring a groundbreaking context window of up to one million tokens.
What the expanded context window means
The new token limit allows developers to feed GPT‑5 much longer prompts, enabling the generation of extensive documents, detailed codebases, and multi‑chapter narratives without needing to truncate or split inputs.
Compared with GPT‑4’s 8,000‑token limit, the million‑token window represents a more than 100‑fold increase, opening possibilities for complex tasks such as full‑book drafting, comprehensive legal analysis, and large‑scale data summarisation.
Key improvements beyond token length
In addition to the enlarged context, OpenAI reports enhancements in reasoning, factuality, and multilingual support. The model is also tuned to reduce hallucinations and to follow user instructions more reliably.
- Improved chain‑of‑thought reasoning
- Higher fidelity in code generation
- Better handling of rare languages
Implications for developers and enterprises
With the ability to process massive inputs, businesses can integrate GPT‑5 into workflows that previously required custom pipelines, such as processing entire research reports, aggregating customer feedback, or generating comprehensive marketing plans in a single request.
OpenAI also announced updated pricing tiers to reflect the increased computational resources required for million‑token operations, though exact figures were not disclosed at launch.
“The million‑token context window fundamentally changes how we think about AI‑assisted creation,” said Mira Murati, OpenAI’s chief technology officer.
Developers interested in testing the new capabilities can access GPT‑5 through the existing API platform, with early‑access credits available for qualified partners.