On 7 August 2025, OpenAI launched its long-awaited GPT-5 model. The model is immediately available to all 700 million weekly ChatGPT users, including users of the free version. During a press briefing Sam Altman referred to GPT-5 as a significant step towards AGI, a model that feels like you are talking to a Ph.D. expert.
In this update, we explore what GPT-5 means in practice, how it differs from previous models, and how companies can take advantage of this new generation of AI. We will also briefly review the latest developments at competitors and conclude with concrete tips for organizations.
What’s new about GPT-5?
Integrated model and smart router
GPT-5 is the first ‘unified’ model from OpenAI. It combines the quick responsiveness of the GPT series with the step-by-step reasoning of the o-series. ChatGPT no longer needs to offer different models; a real-time router determines whether a question can be answered immediately or if extra thinking time is needed. This way, the user always gets the right balance between speed and depth without changing settings.
Variants and subscriptions
OpenAI introduces three variants of GPT-5: gpt-5, gpt-5-mini, and gpt-5-nano. Mini and nano work faster or cheaper and are primarily intended for API use. Free ChatGPT users have access to GPT-5 and mini, while Plus subscribers enjoy higher limits. Pro subscribers (approx. €180 per month) have unlimited access, plus GPT-5-pro and GPT-5-thinking, variants that utilize more computing power for deeper answers. API rates start at $1.25 per million input tokens and $10 per million output tokens.
Larger memory and context window
GPT-5 can process 256,000 tokens of context, a significant improvement over the 200,000 tokens in o3. This means that the model can analyze entire books or large codebases without losing track.
Improved reasoning capacity and test-time compute
In addition to the integrated router, GPT-5 features a new ‘test-time compute’ function: when a question is complex, the model automatically uses extra computing power to think longer. Sam Altman described GPT-5 as a model that feels as if you are talking to a PhD expert. In benchmark tests, GPT-5 outperforms previous models on programming tasks (74.9 % on SWE-Bench Verified) and on health questions.
Fewer hallucinations and safer output
OpenAI reports that GPT-5 reduces hallucinations by 26 % compared to GPT-4o and by 65 % compared to o3. The Thinking variant can make its own reasoning chain visible and conducts 5,000 hours of additional red-team testing to limit misuse. The model uses safe suggestions (‘safe completions’) for risky prompts.
Multimodality, personalization, and new features
OpenAI builds further on the multimodality of GPT-4o. Pro users will soon be able to link their Gmail, Google Contacts, and Calendar; ChatGPT will automatically reference this data when relevant. Users can also select a chat color and choose from four new personalities (Cynic, Robot, Listener, and Nerd) that influence response behavior. Additionally, Canvas (a workspace for diagrams and workflows) is coming to GPT-5, and Voice functionalities are being expanded (such as speech-driven input and real-time conversations).
Software on demand and AI agents
A key theme of GPT-5 is software on demand: the model can build a fully functioning web app from just a few lines of instruction. Furthermore, GPT-5 excels in agent tasks such as executing long tool chains, planning, writing code, and autonomously retrieving information. These improved agency skills extend from the o-series and allow companies to automate advanced workflows.
What can you do now as a company?
Start experimenting immediately
Don't wait for GPT-6. By starting now with GPT-5 and other leading models, you build experience with prompt engineering, chain-of-thought, and multimodality. Start with small-scale pilots: build an internal chat assistant, let AI generate automatic reports, or set up a RAG system to combine internal and external data.
Map processes and data
Inventory which business processes you want to automate. Consider customer service, reporting, document analysis, or code generation. Analyze what data you have and ensure good structure, quality, and compliance with privacy legislation. You will make the most of a large context window like that of GPT-5 when your datasets are well organized.
Develop AI agents and integrations
The combination of reasoning and multimodality makes AI agents more powerful than ever. Use frameworks like LangChain or CrewAI to build agents that autonomously perform tasks, from retrieving data to generating code. Integrate existing tools such as CRM systems, email, and calendars so that agents fit into the daily workflow.
Invest in talent and governance
Ensure staff gain knowledge about prompt engineering, data governance, and AI ethics. Involve legal counsel to comply with the EU AI Act and other regulations. Implement guidelines for safe model deployment and prevent hallucinations and bias through good instructions and monitoring.
Monitor the market and keep comparing
GPT-5 sets a new standard, but the competition is catching up quickly. Stay informed about improvements in Claude 4, Gemini 2.5 Pro, Medium 3, and future releases. These models differ in price, context window, speed, and reasoning; a smart strategy leverages multiple models for different tasks.
The role of MSTR as an AI partner
As an AI partner, MSTR helps organizations implement GPT-5 and other modern LLMs. Our experts closely follow developments at OpenAI, Anthropic, Mistral, and Google and build custom solutions that adapt with new releases. We analyze your processes, establish secure data flows, and develop AI agents that connect with your data. With our modular approach, you can easily upgrade to GPT-6 or other frontier models later.
Conclusion GPT-5
GPT-5 marks an important milestone in the evolution of generative AI. With an integrated reasoning system, a context window of 256,000 tokens, improved accuracy, multimodal functionality, and powerful agents, it offers unprecedented opportunities for businesses.
The right moment to take action is now. Starting AI experiments and mapping processes today builds a solid foundation on which you can easily connect GPT-5 later. MSTR assists you at every step with expertise, customization, and a forward-looking perspective. Want to know what GPT-5 and other LLMs could mean for your business? Feel free to reach out for a no-obligation consultation.
Co-founder