GPT-5 - OpenAI's latest and most powerful AI model

What is GPT-5?

GPT-5 is OpenAI's latest AI model, currently the most powerful model available, and open to all users. GPT-5 is a unified system comprising a base model for answering common questions, a deep reasoning model (GPT-5 Thinking Module) for handling complex problems, and a real-time routing module that intelligently schedules models based on dialogue type, question complexity, tool requirements, and user instructions (such as "think carefully"). GPT-5 excels in multiple fields including programming, writing, mathematics, and health, significantly reducing illusions and errors, and providing answers closer to reality. GPT-5 supports multimodal input and can process both images and text. GPT-5 offers Free, Plus, and Pro modes to meet the needs of different users.

OpenAI recently released a new version of GPT-5 (gpt-5-oct-3) to better address mental health issues. The new version, developed in collaboration with over 170 medical experts worldwide, optimizes responses to delusions, suicidal tendencies, and emotional dependence, and adds features to guide users to seek professional help. The updated model significantly improves safety, reducing negative responses by 39% to 52% and maintaining over 95% reliability in long conversations.

Main functions of GPT-5

• Efficient answering and reasoningIt supports automatically switching between the base model and the deep reasoning model based on the complexity of the problem, quickly answering common questions or delving into complex problems.
• Programming skillsGPT-5 is OpenAI's most powerful programming model, capable of generating high-quality code, debugging large codebases, and performing exceptionally well in production-level tasks.
• Writing and Creative Expression: Transforms rough ideas into compelling writing with literary depth and rhythm, supports various genres, and provides creative inspiration.
• Health ConsultationThey perform better on health-related issues, providing accurate advice, proactively pointing out potential problems, and ensuring that their answers are safe and reliable.
• Multimodal understandingIt supports image and video input, can perform visual reasoning and spatial analysis, and handle multimodal tasks.
• Reduce hallucinations and improve accuracySignificantly reduces hallucination rates, decreases factual errors, and results in more accurate and honest answers.
• Safety and ReliabilityIntroducing a "safe completion" training method reduces deception and ensures safety in high-risk fields such as biology and chemistry.
• Customization and personalizationIt offers four preset personalities (cynic, robot, listener, and nerd), and users can choose the interaction style according to their needs.
• Multilingual supportIt supports multilingual dialogue, can understand and generate multilingual content, and helps users with language conversion.
• API supportGPT-5 offers a variety of model options (such as GPT-5, GPT-5 mini, and GPT-5 nano) through its API, with flexible pricing, making it convenient for developers to use.
• gpt-5-oct-3:
  • For delusions or mental illnessThe model expresses sympathy, does not affirm unfounded fantasies, and guides users to face reality.
  • Targeting suicidal tendenciesThe model can respond to suicide-related conversations more safely and guide users to contact professional crisis hotlines.
  • Targeting emotional dependenceThe model encourages users to build real-world relationships and reminds them to take a break after long conversations.

GPT-5 version information

• GPT-5: The default version, suitable for most general tasks, can automatically switch between basic model and deep inference mode according to the complexity of the problem.
• GPT-5 MiniA smaller, faster version, suitable for lightweight tasks or for continued use after reaching usage limits.
• GPT-5 NanoMinimal version, designed for developers, suitable for rapid prototyping and efficient handling of lightweight tasks.
• GPT-5 Pro: Premium version, available for Pro subscribers, with more powerful computing resources, suitable for complex tasks and deep inference.

GPT-5 performance

• Programming and toolchain capabilities:
  • SWE-bench Verified:74.9% (GPT-4: 52%, o3: 69.1%)
  • Aider Polyglot88%, error rate 33% lower than o3.
  • Front-end developmentInternal test win rate: 70%
  • τ²-bench toolchain tasks96.7%
• Mathematics and Multimodal Capabilities:
  • AIME 2025 Mathematics AssessmentPro+Python mode 100%
  • MMMU Multimodal Understanding84.2%
• professional field:
  • HealthBench Hard (Medical)46.2%
• Accuracy and reliability of knowledge:
  • Error rateApproximately 45% lower than GPT-4o
  • thinking patternsIt is about 80% lower than O3.
  • Hallucination rateOnly 1/6 of o3
  • Deception rate 2.1% (o3 is 4.8%)
• Human-computer interaction and style:
  • Flattery tendenciesSycophancy decreased to 6% (GPT-4 was 14.5%).

GPT-5 official website address

• Official website addressChatGPT

How to use GPT-5

• Visit the official websiteVisit the official GPT-5 website, ChatGPT.
• Free users:
  • Access methodsGPT-5 is now available globally, and free users can use it directly in ChatGPT.
  • limitFree users will have significantly limited GPT-5 usage. Once the usage is exhausted, the model will be switched to GPT-5 mini.
• Plus usersPlus users have a higher GPT-5 usage limit.
  • Additional featuresPlus users can use GPT-5 in Codex CLI to enable "Thinking Mode".
• Pro usersPro users have unlimited access to GPT-5 and can use GPT-5 Pro.
  • FunctionPro users have access to all GPT-5 features, including advanced programming and authoring tools.
• Enterprise and Education Edition usersEnterprise and Education edition users will gain access within one week.
• Using GPT-5 in the API:
  • API calling method:
    • Supported modelsThe GPT-5 models supported in the API include gpt-5, gpt-5-mini, gpt-5-nano, and gpt-5-chat.
    • Input/OutputSupports inputting images and text, and outputting text.
    • EndpointSupports endpoints such as /chat/completions, /responses, and /assistants.
  • Function support:
    • StreamingSupports streaming output.
    • Function callingSupports function calls.
    • Structured outputsSupports structured output.
    • Fine-tuningSupports model fine-tuning.
    • Responses APISupports features such as web search, file search, code interpreter, and MCP.

GPT-5 product pricing

• Free usersAll users can use GPT-5 for free, but there is a limit to the number of times it can be used.
• Plus users$20 USD/month, with a higher usage limit, suitable for frequent daily use.
• Pro users$200 USD/month for unlimited access and the ability to use the GPT-5 Pro version.
• API pricing:
  • GPT-5For every million input tokens, $1.25 is awarded, and $10 is awarded in output.
  • GPT-5 MiniFor every million input tokens, $0.25 is generated, and $2 is generated.
  • GPT-5 NanoFor every million input tokens, $0.05 is awarded, and $0.40 is awarded in output.

Application scenarios of GPT-5

• Programming DevelopmentIt generates high-quality code, debugs large codebases, and performs well in production-level tasks, making it suitable for software development, front-end design, and complex system development.
• Writing and CreationGPT-5 can transform rough ideas into well-written pieces with literary depth and rhythm, supporting a variety of genres and suitable for writing articles, reports, poems, and creative writing.
• Health ConsultationIt provides accurate advice, proactively alerts potential problems, and ensures that the answers are safe and reliable, making it suitable for health consultations and medical science popularization.
• Multimodal tasksIt performs visual reasoning and spatial analysis, and is suitable for handling chart interpretation, photo summarization, and multimodal tasks.
• Complex Problem SolvingIt is suitable for handling complex scientific problems, mathematical puzzles, and multi-step tasks, and is applicable to scientific research, advanced education, and professional consulting.

Official examples of GPT-5

Rolling Ball Game

Prompt:Create a single-page app in a single HTML file with the following requirements:

• Name: Jumping Ball Runner
• Goal: Jump over obstacles to survive as long as possible.
• Features: Increasing speed, high score tracking, retry button, and funny sounds for actions and events.
• The UI should be colorful, with parallax scrolling backgrounds.
• The characters should look cartoonish and be fun to watch.
• The game should be enjoyable for everyone.

(Chinese) prompt wordsCreate a single-page application using only one HTML file that meets the following requirements:

• Game Name: Jump Ball Parkour
• Game objective: Jump and avoid obstacles to survive as long as possible.
• Game features: gradually increasing speed, record high score, retry button, and fun sound effects for actions and events.
• User interface: colorful with a parallax scrolling background.
• Character design: The characters should have a cartoon style and look fun.
• Game experience: Suitable for everyone.

Pixel Art

Prompt：Create a single-page app, in a single HTML file, that provides a retro pixel painting experience.

• Canvas: fixed pixel grid with zoom; tools for pencil, eraser, fill, line, rectangle, circle; grid toggle.
• Palette: 16-color swatches with two custom slots; eyedropper; foreground/background swap.
• Editing: undo/redo, copy/paste selection, flip/rotate selection, clear canvas; status bar with cursor coords.
• UI shell: faux OS window (’90s style) with draggable title bar, toolbar icons, tooltip hints.
• Import/Export: import PNG (quantize to palette) and export PNG/SpriteSheet + JSON; save/load from localStorage.
• Shortcuts: number keys for tools, +/- for zoom; accessible labels and focus order.
• Responsive layout; no uploads to servers.