OpenAI releases the long-awaited GPT-5 a few days after announcing their first open-weight model in six years. What’s even better, you can Start using it noweven if you are a free user. The company claims that GPT-5 is the best model yet in terms of coding, writing and accuracy. OpenAI CEO Sam Altman (and hypeman) said
“GPT-5 is the first time that it really feels like you’re talking to an expert in any topic,” during a company press briefing held before today’s official announcement. “It reminds me of when the iPhone went from those giant, old pixel [screens] to the Retina Display, and then I went back to using one of those big pixelated things and I was like, ‘Wow, I can’t believe how bad we had it.'”
Altman promised at the beginning of the year that GPT-5 would provide a unified user experience. The new model delivers on this promise. OpenAI’s default model is now a reasoning one, which means that the system is programmed in a way to break down complex problems into smaller pieces. Previously, to force ChatGPT’s reasoning model to be used, you had select the “Think Longer” from the prompt bar. Most free users were unaware that OpenAI had better models. The company has simplified ChatGPT with GPT-5.
There are only three versions available to consumers. GPT-5 Mini is only available to free and Plus users who have reached their GPT-5 usage limits. GPT-5 Pro is only available to those who subscribe to the $200 per month Pro plan. Plus users have more query limits than those with free accounts, while Pro customers are allowed to chat with GPT-5 at any time.
A graph highlighting some of GPT-5’s enhanced capabilities
(Openai)
GPT-5 is faster at reasoning than OpenAI’s previous AI, o3, which was the previous state-ofthe-art AI. Altman said “It’s so fast that I’ve had the psychological experience of wondering, like, is it really thinking enough? And then it gives a great answer,” . OpenAI claims that the model is more accurate than its previous reasoning systems, and suffers from less hallucinations. GPT-5, for example, is about 80 percent less likely than o3 to make a factual mistake in its answer when thinking. We’ll have to wait and see how GPT-5 performs in real-world situations, but if OpenAI made significant improvements, this would be a huge deal. Hallucinations are a major weakness for reasoning models, especially when compared to their large language counterparts.
OpenAI claims GPT-5 to be its safest AI yet. It includes a feature called Safe Completions. Alex Beutel is the safety research lead at OpenAI. “This worked well, but as a challenge that there can be kind of carefully worded prompts that could be confusing. So if someone says how much energy is needed to ignite some specific material that could be an adversary trying to get around the safety protections and cause harm, or it could be a student asking a science question to understand the physics of this material.”
Safe Completions will allow GPT-5 to provide the most helpful answers within the safety constraints that OpenAI has placed on it. In situations such as the one Beutel described above, the model only provides high-level information which cannot be used to harm anyone. Beutel continued.
GPT-5 is also better at identifying concerns and suggesting questions that the user should ask their healthcare provider when it comes health-related questions. It will also respond to those prompts with greater precision, thanks to its ability to adapt to a person’s knowledge and geography. OpenAI claims that GPT-5, on top of all else, is its best coding model yet. The chatbot can also be a better writer, as the company claims it is better at translating drafts into “compelling, resonant” text.
OpenAI will be adding new features to ChatGPT along with GPT-5. Users can now select a color to use for their chats. Some options are exclusive to paying customers. OpenAI has made it easier for ChatGPT users to connect to Gmail and Google Calendar. Once you’ve enabled the connections, your chatbot will automatically refer to your Google accounts. You won’t have to select anything to start chatting. OpenAI will roll out this feature for Pro subscribers next week. Other users will follow. OpenAI has introduced a few pre-set personalities in the Custom Instructions Pane, where users can write system prompts that will change how ChatGPT responds to them. The four options – cynics, robots, listeners and nerds – are available in a research preview and can be changed at any time.
OpenAI has released an updated version to its Advanced Voice feature, which was introduced last summer. OpenAI claims that the tool can better understand instructions and adapt its speaking style according to the situation. OpenAI retires Standard Voice Mode as part of this change. In practice, this means that the company can offer a better experience to all users since it no longer needs to rely on Standard Voice Mode. Standard Voice Mode is not natively multimodal like Advanced Voice, and therefore does not understand the nuances of the human voice. Altman’s response to the question of where OpenAI stands on the path towards artificial general intelligence is
. “I kind of hate the term AGI, because everyone at this point uses it to mean a slightly different thing, but [GPT-5] is a significant step forward towards models that are really capable. We’re still missing something quite important,” Altman said, noting that GPT-5 cannot continuously learn on its behalf. “But the level of intelligence here, the level of capability, it feels like a huge improvement. Certainly, if I could go back five years before GPT-3 and you told me we have this now, I’d be like that’s a significant fraction of the way to something very AGI-like.”
Updated 2:00PM: Added context about hallucination rate.
