OpenAI’s GPT-5 cuts hallucinations by 80% while reaching 700M users

OpenAI has launched GPT-5 and three variants—GPT-5 Pro, GPT-5 mini, and GPT-5 nano—making its latest AI system available to all ChatGPT users, including free tier subscribers for the first time. The release marks OpenAI’s attempt to unify its AI capabilities into a single system with reduced hallucinations, improved coding performance, and a new “safe completions” approach that provides helpful responses within safety boundaries rather than outright refusals.

What you should know: GPT-5 introduces a unified system architecture that automatically routes queries between different processing approaches based on complexity and user needs.

• The system combines a smart, efficient model for most questions with a deeper reasoning model called “GPT-5 thinking” for harder problems.
• Like GPT-4o, GPT-5 is multimodal, supporting text, image, and voice interactions.
• The rollout extends to ChatGPT’s 700 million weekly active users with different usage limits based on subscription tiers.

Technical improvements: OpenAI claims significant performance gains across multiple benchmarks, positioning GPT-5 as its “strongest coding model yet.”

• The model achieves 74.9 percent on SWE-bench Verified and 88 percent on Aider Polyglot coding benchmarks.
• For health-related queries, GPT-5 scores 46.2 percent on HealthBench Hard, though OpenAI emphasizes it “does not replace a medical professional.”
• Mathematics performance reaches 94.6 percent on AIME 2025 without tools, while multimodal understanding hits 84.2 percent on MMMU.

Reduced hallucinations: OpenAI reports substantial improvements in factual accuracy compared to previous models.

• With web search enabled, GPT-5’s responses are about 45 percent less likely to contain factual errors than GPT-4o.
• When using “thinking” mode, the model shows about 80 percent fewer factual errors than o3.
• On long-form content benchmarks, GPT-5 with thinking demonstrates about six times fewer hallucinations than o3.
• The company has also reduced sycophantic replies—overly agreeable responses—from 14.5 percent to less than 6 percent through improved training.

New features and interface updates: ChatGPT receives several user experience enhancements alongside the model upgrade.

• Customizable chat colors and preset conversation personalities with options like “Cynic,” “Robot,” “Listener,” and “Nerd.”
• Integration with Gmail, Google Calendar, and Google Contacts for Pro users.
• A unified “Advanced Voice” system that better understands instructions and adapts speaking style.

Developer access and pricing: The API release includes three versions with different performance and cost tradeoffs.

• GPT-5 API pricing lands at $1.25 per million input tokens with a 90 percent cache discount and $10 per million output tokens.
• GPT-5 Mini offers a more economical option at $0.25 input/$2 output per million tokens.
• GPT-5 Nano provides the most cost-effective tier at $0.05 input/$0.40 output per million tokens.

Competitive landscape: The launch intensifies competition as OpenAI faces pressure from rivals across the AI industry.

• Earlier this week, Anthropic released Claude Opus 4.1, which scored 74.5 percent on SWE-bench compared to GPT-5’s 74.9 percent.
• OpenAI reports having 5 million paying business users and 4 million developers building on its API platform.
• The company competes with Google’s Gemini models, Anthropic’s Claude family, and Meta’s open-weight Llama models.

Rollout details: GPT-5 begins replacing multiple existing models as the default option for ChatGPT users.

• The model replaces GPT-4o, OpenAI o3, OpenAI o4-mini, GPT-4.1, and GPT-4.5 as the default for signed-in users.
• Enterprise and education customers receive access next week following the initial consumer rollout.
• Free users transition to GPT-5 mini after reaching their usage limits for the full model.