OpenAI Launches GPT-5

<p>OpenAI has launched GPT‑5&period; The company says GPT‑5 is a significant leap in intelligence over all its previous models&comma; featuring state-of-the-art performance across coding&comma; math&comma; writing&comma; health&comma; visual perception&comma; and more&period; It is a unified system that knows when to respond quickly and when to think longer to provide expert-level responses&period; GPT‑5 is available to all users&comma; with Plus subscribers getting more usage&comma; and Pro subscribers getting access to GPT‑5 pro&comma; a version with extended reasoning for even more comprehensive and accurate answers&period;<&sol;p>&NewLine;<p>CEO Sam Altman announced during the product introduction that Chat GPT has over 700 million users&period;<&sol;p>&NewLine;<p>GPT‑5 is a unified system with a<b> smart&comma; efficient model<&sol;b> that answers most questions&comma; a <b>deeper reasoning model<&sol;b> &lpar;GPT‑5 thinking&rpar; for harder problems&comma; and a <b>real‑time router<&sol;b> that quickly decides which to use based on conversation type&comma; complexity&comma; tool needs&comma; and your explicit intent &lpar;for example&comma; if you say &OpenCurlyDoubleQuote;think hard about this” in the prompt&rpar;&period; The router is continuously trained on real signals&comma; including when users switch models&comma; preference rates for responses&comma; and measured correctness&comma; improving over time&period; Once usage limits are reached&comma; a mini version of each model handles remaining queries&period; In the near future&comma; we plan to integrate these capabilities into a single model&period;<&sol;p>&NewLine;<p>GPT‑5 is the strongest coding model to date&period; It shows particular improvements in <b>complex front‑end generation<&sol;b> and <b>debugging larger repositories<&sol;b>&period; It can often create beautiful and responsive websites&comma; apps&comma; and games with an eye for aesthetic sensibility in just one prompt&comma; intuitively and tastefully turning ideas into reality&period; Early testers also noted its design choices&comma; with a much better understanding of things like spacing&comma; typography&comma; and white space&period; <a class&equals;"transition ease-curve-a duration-250 text-primary-100 hover&colon;text-primary-60 relative underline-offset-&lbrack;0&period;25rem&rsqb; decoration-1 underline" href&equals;"https&colon;&sol;&sol;openai&period;com&sol;index&sol;introducing-gpt-5-for-developers&sol;">See here<&sol;a> for full details on what GPT‑5 unlocks for developers&period;<&sol;p>&NewLine;<p>GPT‑5 is significantly less likely to hallucinate than previous models&period;  With web search enabled on anonymized prompts representative of ChatGPT production traffic&comma; GPT‑5’s responses are ~45&percnt; less likely to contain a factual error than GPT‑4o&comma; and when thinking&comma; GPT‑5’s responses are ~80&percnt; less likely to contain a factual error than OpenAI o3&period;<&sol;p>&NewLine;