OpenAI has officially launched GPT-5, calling it the smartest and most capable version of its artificial intelligence models to date. The company says the new model is more accurate, more efficient, and has significantly lower hallucination rates compared to previous versions.
Taking to X, OpenAI CEO Sam Altman posted: “GPT-5 is the smartest model we've ever done, but the main thing we pushed for is real-world utility and mass accessibility/affordability.”
Who can access GPT-5?
GPT-5 is now being gradually rolled out to ChatGPT Plus, Pro users, and Team plans across web, mobile, and desktop platforms. It will soon be available to Enterprise and Education plan users as well.
Anyone using ChatGPT can access GPT-5. If you’re on a paid plan (Plus, Pro, or Team), you can also use the model picker to switch between the regular GPT-5 model and GPT-5 Thinking.
Users on the Pro and Team tiers get access to GPT-5 Thinking Pro, which takes a little longer to respond but offers more accurate results for complex tasks.
Usage limits by tier
Free users can send up to 10 messages every 5 hours, after which ChatGPT switches to a smaller version of the model until the limit resets. Free users can also send one GPT-5 Thinking message per day.
Plus users have a higher limit of 80 messages every 3 hours. Thereafter, their conversations also default to a lighter version of GPT-5. They can use GPT-5 Thinking up to 200 times per week. Once this limit is reached, GPT-5 Thinking disappears from the picker, although automatic switches by the system won’t count toward this limit.
Pro and Team subscribers enjoy unlimited access to GPT-5 models, with appropriate safeguards in place to prevent misuse, according to the company.
A unified AI system
GPT-5 introduces a more intelligent system that simplifies how the model works behind the scenes. Instead of users needing to choose between versions such as Turbo or Standard, the system now automatically selects the best fit based on the task. There’s a fast, general-purpose model for everyday queries, and GPT-5 Thinking for more advanced reasoning. A built-in router decides which to use based on the complexity of the conversation and your instructions.
Once usage limits are hit, a mini version takes over, ensuring a smooth experience even when you’ve reached your cap. In the future, OpenAI plans to combine all of these capabilities into a single seamless model.
Integration with Google apps
A major new feature is the ability to connect ChatGPT to Gmail, Google Calendar, and Google Contacts. Once linked, ChatGPT can use information from these apps without needing repeated prompts.
Personalisation with accent colours
ChatGPT now supports accent colour customisation, allowing users to personalise the look of chat bubbles, highlighted text, and the Voice button. This can be done through the Settings menu on both web and mobile apps.
Smarter Voice mode
OpenAI is rolling out a more advanced Voice Mode, now compatible with custom GPTs (user-created for a specific purpose). This replaces the old "Standard Voice Mode", which will be retired in 30 days. Plus users will get near-unlimited use, while Free users can use it for several hours daily.
Voice interactions will now adapt to your preferences, changing tone, speed, and speaking style depending on your instructions or the context.
New ChatGPT personalities
You can now choose from five ChatGPT personalities in the settings, including the default neutral style. Options include:
However, these do not apply in Voice Mode.
‘Strongest coding model’
OpenAI describes GPT-5 as “our strongest coding model to date.” On its website, the company demonstrates how GPT-5 was able to build an entire game from a single prompt.
It outperforms earlier models such as o3 across coding benchmarks and has been tuned for agentic coding products such as GitHub Copilot, Cursor, Codex CLI, and others.
Older models being retired
With the rollout of GPT-5, several older models are being phased out. These include GPT-4o, GPT-4.1, GPT-4.5, and all o3 and o4-mini variants. If you open an older conversation that used one of these models, ChatGPT will automatically upgrade it to the nearest GPT-5 version.
For example, chats that used o3 will now run on GPT-5 Thinking, while those using o3-Pro will run on GPT-5 Pro (available only on Pro and Team plans). Pro, Team, and Enterprise users will still have access to legacy models through special settings for a limited time.
Taking to X, OpenAI CEO Sam Altman posted: “GPT-5 is the smartest model we've ever done, but the main thing we pushed for is real-world utility and mass accessibility/affordability.”
Who can access GPT-5?
GPT-5 is now being gradually rolled out to ChatGPT Plus, Pro users, and Team plans across web, mobile, and desktop platforms. It will soon be available to Enterprise and Education plan users as well.
Anyone using ChatGPT can access GPT-5. If you’re on a paid plan (Plus, Pro, or Team), you can also use the model picker to switch between the regular GPT-5 model and GPT-5 Thinking.
Users on the Pro and Team tiers get access to GPT-5 Thinking Pro, which takes a little longer to respond but offers more accurate results for complex tasks.
Usage limits by tier
Free users can send up to 10 messages every 5 hours, after which ChatGPT switches to a smaller version of the model until the limit resets. Free users can also send one GPT-5 Thinking message per day.
Plus users have a higher limit of 80 messages every 3 hours. Thereafter, their conversations also default to a lighter version of GPT-5. They can use GPT-5 Thinking up to 200 times per week. Once this limit is reached, GPT-5 Thinking disappears from the picker, although automatic switches by the system won’t count toward this limit.
Pro and Team subscribers enjoy unlimited access to GPT-5 models, with appropriate safeguards in place to prevent misuse, according to the company.
A unified AI system
GPT-5 introduces a more intelligent system that simplifies how the model works behind the scenes. Instead of users needing to choose between versions such as Turbo or Standard, the system now automatically selects the best fit based on the task. There’s a fast, general-purpose model for everyday queries, and GPT-5 Thinking for more advanced reasoning. A built-in router decides which to use based on the complexity of the conversation and your instructions.
Once usage limits are hit, a mini version takes over, ensuring a smooth experience even when you’ve reached your cap. In the future, OpenAI plans to combine all of these capabilities into a single seamless model.
Integration with Google apps
A major new feature is the ability to connect ChatGPT to Gmail, Google Calendar, and Google Contacts. Once linked, ChatGPT can use information from these apps without needing repeated prompts.
Personalisation with accent colours
ChatGPT now supports accent colour customisation, allowing users to personalise the look of chat bubbles, highlighted text, and the Voice button. This can be done through the Settings menu on both web and mobile apps.
Smarter Voice mode
OpenAI is rolling out a more advanced Voice Mode, now compatible with custom GPTs (user-created for a specific purpose). This replaces the old "Standard Voice Mode", which will be retired in 30 days. Plus users will get near-unlimited use, while Free users can use it for several hours daily.
Voice interactions will now adapt to your preferences, changing tone, speed, and speaking style depending on your instructions or the context.
New ChatGPT personalities
You can now choose from five ChatGPT personalities in the settings, including the default neutral style. Options include:
- Cynic: Sarcastic and blunt, but helpful.
- Robot: Direct and emotionless.
- Listener: Calm, thoughtful, and warm.
- Nerd: Curious and loves explaining things.
However, these do not apply in Voice Mode.
‘Strongest coding model’
OpenAI describes GPT-5 as “our strongest coding model to date.” On its website, the company demonstrates how GPT-5 was able to build an entire game from a single prompt.
It outperforms earlier models such as o3 across coding benchmarks and has been tuned for agentic coding products such as GitHub Copilot, Cursor, Codex CLI, and others.
Older models being retired
With the rollout of GPT-5, several older models are being phased out. These include GPT-4o, GPT-4.1, GPT-4.5, and all o3 and o4-mini variants. If you open an older conversation that used one of these models, ChatGPT will automatically upgrade it to the nearest GPT-5 version.
For example, chats that used o3 will now run on GPT-5 Thinking, while those using o3-Pro will run on GPT-5 Pro (available only on Pro and Team plans). Pro, Team, and Enterprise users will still have access to legacy models through special settings for a limited time.