OpenAI releases GPT-5: its smartest and fastest AI model to date
OpenAI introduced its latest large language model last week: GPT-5. According to the company, it is their most intelligent, fastest, and most useful model to date. CEO and co-founder Sam Altman describes GPT-5 as a PhD-level expert who can discuss virtually any topic.
GPT-5 requires less complex prompts and additional context to arrive at a good answer. In demonstrations, the model shows how it can break down and analyse compound questions. Moreover, GPT-5 indicates when it doesn’t know something, instead of making up a fictional answer.
The model also excels in mathematics and programming, and users can adjust the assistant’s tone, from concise to friendly or even sarcastic.
One of the most impressive features of GPT-5 is its greatly improved capability in software development. The model can generate complete applications based on just a single request – with thoughtful layouts, responsive designs, and typography that is visually convincing as well. OpenAI itself calls this vibe coding. It also excels at debugging complex codebases, making it a powerful assistant for developers at any level.
OpenAI is immediately releasing several variants: in addition to the standard model, there are also GPT-5 mini and nano, available via the API and accessible to free users.
For the first time, free users get access to the same model as Plus users, albeit with more limited usage limits. Once those limits are reached, they automatically switch to a ‘mini’ version of the model. The same applies to Plus users, although their limits are more generous.
The more powerful GPT-5 pro and thinking versions are exclusive to Pro subscribers, who pay $200 a month. These advanced models can be linked to Gmail, Google Calendar, and Google Contacts, allowing them to manage appointments, for example.
OpenAI demonstrated that the new models outperform their predecessors on various benchmarks, including SWE-bench Verified for software engineering, Aider Polyglot for multilingual code editing, and MMMU for visual problem-solving. Still, the performance improvements are generally less spectacular than with previous GPT upgrades.
OpenAI has decided to phase out all its other models. As a result, previous conversations with older models are automatically converted to the most suitable GPT-5 equivalent. Specifically, this means conversations with 4o, 4.1, 4.5, 4.1-mini, o4-mini, or o4-mini-high will open in GPT-5, conversations with o3 will open in GPT-5 Thinking, and conversations with o3-Pro will open in GPT-5 Pro (only available for Pro and Team subscriptions).
The so-called voice mode will, for now, continue to run on GPT-4o, according to OpenAI. An interesting innovation is that voice mode now also works with custom GPTs. Paid users can give specific instructions regarding things like answer length, speaking speed, tone, and more.
Microsoft has already announced that it will integrate GPT-5 into Microsoft 365 Copilot, GitHub Copilot, and Azure AI Foundry.
Pro users will also be able, starting next week, to link their Google Calendar, Google Contacts, and even their Gmail account to ChatGPT, to receive more personalized assistance within the chat environment.
During the launch event, a demo was shown in which an OpenAI engineer used ChatGPT to organise his schedule. The model proactively helped to fit in a running session between appointments.
OpenAI also announced that it will offer its ChatGPT Enterprise product to US federal agencies for just $1 over the coming year. This makes the technology available to employees of the federal executive branch at virtually no cost.
Agencies will have access to the latest models via ChatGPT Enterprise, including features such as Advanced Voice Mode (for an additional period of 60 days). OpenAI will open its first office in Washington DC early next year to work more closely with policymakers and regulators.
In June, OpenAI already launched a special offer for government institutions and received a contract worth up to $200 million from the US Department of Defense.
Emerce
Subscribers 0
Fans 0
Followers 0
Followers