On Thursday, OpenAI unveiled its latest artificial intelligence models, known as GPT-5, which promises to deliver substantial enhancements in various domains including coding, healthcare, mathematics, and visual recognition. The San Francisco-based tech firm announced that this cutting-edge model integrates improvements across multiple applications compared to its predecessors. For the first time, GPT-5 will be accessible to all users, encompassing all subscription levels and individuals utilizing the free tier.
GPT-5 Plans and Availability
According to a blog post by OpenAI, the rollout of GPT-5 will reach all users, including ChatGPT Plus, Pro, and Team subscribers, as well as those on the free plan. Users in the Enterprise and Education tiers can expect access to the model next week, while paid subscribers will also have the model available in the Codex CLI.
The launch of GPT-5 will replace previous models including GPT-4o, OpenAI o3, OpenAI o4-mini, GPT-4.1, and GPT-4.5 for those logged in. This change aims to streamline the selection process, removing the need for users to choose from multiple models based on tasks. The new model is designed to effectively gauge when to utilize reasoning and when to generate swift responses. Moreover, paid users will have access to a distinct feature called GPT-5 Thinking, which is tailored for in-depth reasoning for complex challenges.
ChatGPT Plus members will enjoy extended access to GPT-5, while the ChatGPT Pro plan, available for $200 (approximately ₹17,500), offers unlimited access along with the added GPT-5 Pro option. The rollout for those in the ChatGPT Enterprise and Education plans will begin on August 14.
OpenAI outlined the pricing structure for GPT-5’s API, which will charge $1.25 (roughly ₹110) per million input tokens and $10 (approximately ₹880) per million output tokens. The GPT-5 Nano variant, the most economical option, offers a cost of $0.05 (around ₹4) for input tokens and $0.40 (approximately ₹35) for output tokens.
GPT-5 Features
The firm asserts that GPT-5 represents a consolidated system of large language models (LLMs) showcasing enhanced efficacy in sectors such as coding, healthcare, mathematics, visual perception, and writing. Furthermore, it is reportedly less susceptible to producing errant information and has improved cognitive awareness, particularly with the GPT-5 Thinking component. This capability allows the system to discern between quick responses and deeper contemplation, ensuring accurate and thorough replies.
OpenAI claims that in various academic and human-evaluated benchmarks, GPT-5 surpasses all earlier models. Specifically in coding, it achieved a score of 74.9 percent in the SWE-bench Verified, considerably higher than the 54.6 percent attained by GPT-4.1, thereby establishing GPT-5 as the most advanced coding model to date.
This generation introduces a real-time router that oscillates between everyday and reasoning modes, consistently trained on real user interactions. OpenAI has indicated that the reduction of hallucinations and an increase in instructional compliance are notable improvements within the model. Additionally, enhanced performance across writing, coding, and health-related tasks has been highlighted.
In the realm of coding, GPT-5 is reported to exhibit capabilities in generating front-end code and debugging larger projects. Users can now request the creation of responsive websites, mobile apps, and games in a single prompt. On the writing front, enhancements allow the model to manage structural ambiguities more effectively, providing a balance between form and expressive quality.
Notably, healthcare has been identified as a crucial area of enhancement. OpenAI describes GPT-5 as functioning like an active thought partner, proactively identifying potential issues and raising inquiries to offer more constructive advice. The model is designed to adapt more effectively to the user’s background, knowledge base, and geographic context, resulting in personalized responses.
Additionally, OpenAI is introducing a research preview of four new preset personalities within the GPT-5-powered ChatGPT. The options—Cynic, Robot, Listener, and Nerd—are customizable experiences that align with users’ communication preferences and are currently applicable for text chat, with future plans to incorporate these into Voice Mode.
OpenAI has reported that it dedicated 5,000 hours to evaluate GPT-5, implementing robust safety protocols based on input from red team execution and internal teams to significantly reduce associated risks. The model has undergone training in a new format known as safe competitions, which emphasizes providing helpful responses while adhering to established safety guidelines. Should the model need to decline a request, it will offer transparent reasoning and suggest safer alternatives, aiming to minimize unnecessary rejections.
Microsoft Incorporates GPT-5
In conjunction with these developments, Microsoft has announced the integration of GPT-5 into its suite of offerings, including Copilot, Microsoft 365 Copilot, and various enterprise and consumer applications. The update enhances the reasoning capabilities of Microsoft 365 Copilot, allowing for longer, more context-aware conversations. According to the company, Copilot is now better equipped to address complex challenges.
Microsoft 365 Copilot license holders can utilize GPT-5 to analyze documents, emails, and files, while Copilot Studio will allow for the creation of custom agents using the model. Users can specifically select GPT-5 in customized prompts.
Furthermore, GitHub plans to incorporate GPT-5 across all paid GitHub Copilot subscriptions, enabling developers to write, edit, and deploy code using GitHub Copilot Chat on platforms like github.com, Visual Studio Code, and GitHub Mobile. Developers working within Visual Studio Code will also benefit from developing agents directly within the code editor.
Finally, users of Azure AI Foundry will gain access to OpenAI’s latest models and leverage the newly introduced router, which optimizes the model selection process based on the complexity, performance requirements, and cost-efficiency of tasks.
