OpenAI has rolled out its newest AI model, GPT‑5, marking a major milestone after two years of anticipation. The GPT‑5 is now powering ChatGPT for all 700 million weekly users, spanning Free, Plus, Pro, and Team tiers, while Enterprise and Education access will follow shortly.
GPT-5’s launch coincides with OpenAI closing an $8.3 billion funding round at a $300 billion valuation, months ahead of schedule and five times oversubscribed, with investors seeking a total of $40 billion. The round was led by Dragoneer Investment Group, which contributed $2.8 billion, and included participation from major private equity firms like Blackstone and TPG, mutual fund managers such as T. Rowe Price, and traditional VCs.
To make room for new strategic partners, OpenAI reduced allocations to early investors, causing some frustration among veteran backers. This positions OpenAI among the most valuable private companies globally, comparable to Coca-Cola, Novo Nordisk, and LVMH. The company’s annual revenue run-rate has grown rapidly, from $10 billion to $13 billion ARR.
CEO Sam Altman described GPT‑5 as the closest yet to an on‑demand PhD‑level expert. He asserts it embodies substantial progress toward artificial general intelligence (AGI), though he emphasised that the path to true AGI with real‑time continuous learning remains unfinished.
The model brings an array of tangible improvements. It is more accurate, less prone to making things up, and far better at tackling complex tasks like coding, writing, health-related queries, and reasoning. Enterprise capabilities like real-time software generation and advanced test‑time compute reflect a clear focus on business utility.
GPT-5 registered benchmark scores of 74.9% on SWE-bench Verified (up from 69.1% for o3), 88% on Aider polyglot coding, and performed at 25.5% on HealthBench Hard, above previous generations. It demonstrated advanced long-context capabilities, handling up to 272,000 input tokens (with total context length of up to 400,000 tokens), and achieved 96.7% on τ-bench telecom, a benchmark where no earlier model scored above 49%.
Here are five things to know about GPT-5!
Available to all ChatGPT Users
GPT-5 is now active across Free, Plus, Pro, and Team users globally. Free users start on GPT-5 but may be switched to a lighter version, “GPT-5 Mini,” during high traffic.
More accurate, less prone to errors
GPT-5 reduces hallucinations by about 45% compared to GPT-4o. It’s better at saying “I don’t know” instead of fabricating answers, and it’s less likely to mirror user bias or agree excessively.
It processes over 2.5 billion prompts daily, and paid business users jumped from 3 million to 5 million in a few months, with overall ChatGPT usage increasing fourfold year-over-year.
Handles complex tasks with depth and memory
Handles complex tasks with depth and memory: GPT-5 excels at executing multi-step tasks, whether it’s building full websites, debugging large codebases, or writing context-rich content. It remembers more within conversations and reasons through problems using a built-in “deep thinking” mode, giving users expert-like assistance.
The model was stress-tested with over 5,000 hours of red team evaluation from external organisations. GPT-5 uses “safe completions,” instead of blanket refusals for risky queries, it offers helpful but securely constrained responses, decreasing deceptive behaviours and “failing gracefully” when necessary.
Enterprise-ready tools and developer flexibility
Enterprise-ready tools and developer flexibility: Businesses can now use GPT‑5 in three versions: GPT‑5, GPT‑5 Mini, and GPT‑5 Nano, each optimised for different use cases. These models bring faster performance, higher accuracy, and cost-efficiency. Developers can access them via API with tiered pricing, making integration into products and services easier.
GPT-5 is also being rolled out across the Microsoft ecosystem, including Microsoft 365 Copilot, GitHub Copilot, Azure AI Foundry, and other Microsoft products, featuring enterprise-grade security, compliance, and privacy via Azure.
Designed for long-term reliability
Designed for long-term reliability: GPT‑5 isn’t just smarter but built with long-term stability in mind. It handles edge cases more gracefully, escalates fewer safety flags, and integrates better with external systems.
OpenAI commands over 70% LLM market share as of January 2024, while competitors like Anthropic are raising billions more at high valuations. The company must convert to a full for-profit structure by December 31, 2025, or risk forfeiting some investment from partners like SoftBank, the transition accompanies executive restructuring, with CEO Sam Altman shifting focus towards research and Brad Lightcap expanding operational duties.
OpenAI is reportedly considering a $500 billion valuation for future employee share sales, possibly setting the stage for an IPO.
Why this matters
GPT‑5 signals OpenAI’s intent to blend technical finesse with real-world utility. This launch cements OpenAI’s market leadership, strategic partnerships, technical advances, and growth, making GPT-5 both a technological and business milestone in the accelerating AI sector.
