{"id":"f-e550251d2f-48ac7051-b085-5577-9ab3-02d6719557e1","title":"OpenAI rolls out GPT-5.4 and GPT-5.4 Pro: What's different","description":"OpenAI has launched GPT-5.4 and GPT-5.4 Pro, introducing stronger reasoning, coding and computer-use abilities for ChatGPT, developers and enterprise workflows.","timestamp":1772960280,"category":"technology","headline":"OpenAI rolls out GPT-5.4 and GPT-5.4 Pro: What's different","modifiedHeadline":"OpenAI Unveils Advanced GPT-5.4 AI Model","subCategories":["scid_b7e786a9"],"subcategoriesName":["Artificial Intelligence"],"region":"IN","bundleMetaData":{"bundleId":"c35cfc57-1bba-4539-a24c-695ea99ac023","bundleTitle":"Mastering AI","cardData":null},"iabCategoryId":"597","fcSummary":"openai has launched gpt-5.4, its newest ai model, designed to enhance professional workflows like coding, research, and presentations. it reduces revisions by planning responses and improves long-context analysis, supporting up to one million tokens. gpt-5.4 also boasts stronger coding capabilities and can now directly operate software, marking a significant step in ai automation. pricing reflects the increased performance.","summary":"OpenAI launches GPT-5.4 & Pro, boosting reasoning, coding & tool use. New features streamline tasks & improve performance.","ctaUrl":"https://trends.glance.com/newz/Daily_Digest/IN/en/cnbc-tv18/f-e550251d2f-48ac7051-b085-5577-9ab3-02d6719557e1","sourceUrl":"https://www.cnbctv18.com/technology/openai-rolls-out-gpt-5-4-and-gpt-5-4-pro-whats-different-ws-l-19863524.htm","bannerImage":"https://g-mob.glance-cdn.com/public/fc/image/V86V5llFab60.webp","content":"OpenAI has introduced GPT-5.4, its latest frontier AI model, now rolling out across ChatGPT, the API and Codex. The company says the release focuses on practical work: coding, research, spreadsheets, presentations and other professional tasks that require longer reasoning and fewer corrections.\n\n
\n
\n The update follows earlier releases in the GPT-5 series and builds directly on GPT-5.2 and GPT-5.3-Codex, combining their reasoning and programming strengths into a single system.\n
\n
What GPT-5.4 does differently\n
\n
\n The model appears in ChatGPT as GPT-5.4 Thinking - a higher-performance option called GPT-5.4 Pro is available for users who need more computing power on complex tasks.\n
\n
\n The main change is how the model approaches real work. OpenAI says GPT-5.4 is designed to complete tasks more directly, reducing the back-and-forth that often happens when AI responses require multiple revisions.\n\n
\n
\n One new feature allowsGPT-5.4 Thinking to outline a plan before producing the final answer. Users can review that plan and adjust instructions while the response is being prepared. The aim is simple - clearer direction, fewer follow-ups.\n
\n
\n The model also improves long-context tasks.GPT-5.4 supports up to one million tokens of context, meaning it can analyse longer documents or multi-step workflows without losing track of earlier instructions.\n
\n
\n Search performance has also been upgraded. On BrowseComp - a benchmark measuring how well AI agents find difficult information online - GPT-5.4 achieved 82.7%, compared with 65.8% for GPT-5.2. The Pro version reached 89.3%, setting a new high score for the test.\n\n
\n
Stronger coding and computer use\n
\n
\n Coding was another focus. GPT-5.4 combines the programming capabilities of GPT-5.3-Codex with broader reasoning features.\n
\n
\n On SWE-Bench Pro, a benchmark for software engineering tasks, GPT-5.4 reached 57.7% accuracy, slightly ahead of GPT-5.3-Codex at 56.8%.\n
\n
\n The model can also operate software directly. OpenAI describes GPT-5.4 as its first general-purpose model with native computer-use capabilities, meaning AI agents can navigate applications, websites and tools while completing tasks.\n
\n
\n Performance on OSWorld-Verified, which tests desktop navigation using screenshots and keyboard or mouse actions, reached 75% accuracy, higher than the 47.3% recorded by GPT-5.2.\n\n
\n
Benchmarks and performance improvement\n
\n
\n Benchmarks focused on professional tasks show similar gains. On GDPval, which measures knowledge-work performance across 44 occupations, GPT-5.4 matched or exceeded industry professionals in 83% of comparisons, compared with 70.9% for GPT-5.2.\n
\n
\n Internal testing also showed improvements in spreadsheet modelling. GPT-5.4 scored 87.3% on tasks similar to work done by junior investment-banking analysts - up from 68.4% for GPT-5.2.\n
\n
Availability, pricing and rollout across ChatGPT and API\n\n
\n
\n GPT-5.4 is now available through the OpenAI API and within ChatGPT. Plus, Team and Pro users can access GPT-5.4 Thinking, which replaces GPT-5.2 Thinking. The earlier model will remain available for several months before being retired.\n\n
\n
Developers can also access GPT-5.4-pro, aimed at more complex workloads.\n
\n
\n API pricing reflects the added capability. OpenAI lists $2.50 per million input tokens for GPT-5.4, compared with $1.75 for GPT-5.2.\n
\n
\n For now, OpenAI describes GPT-5.4 as its most capable model for professional workflows - bringing reasoning, coding and tool-based automation together in a single system.","readDurationInMinutes":"6","author":{"name":"CNBC TV18","logo":"https://glance-mob.glance-cdn.com/public/bubble/jpeg/ecc3d19b-fc52-4a39-881c-5b0936ae723b.jpeg"},"ism":false,"analyticsResponses":[{"measurementId":"383191425","type":"COMSCORE"}],"pointerSummary":["OpenAI released GPT-5.4, enhancing ChatGPT & API for coding, research, and professional tasks.","GPT-5.4 Thinking outlines plans before answering, improving task completion & achieving 89.3% on a key benchmark.","The model now supports 1 million tokens & can directly operate software, with API pricing at $2.50/million tokens."],"tags":["OpenAI","GPT54","GPT54Pro","ChatGPT","ArtificialIntelligence","AIModels","GenerativeAI","CodingAI","EnterpriseAI","AIUpdates","LargeLanguageModels","LLMs","AIProductivity","SoftwareEngineering","API","AIBenchmarks","ContextWindow","ComputerUseAI","GDPval","SWEBenchPro","OSWorldVerified","InvestmentBankingAI"],"consent":true,"redirectionEnabled":false}

OpenAI rolls out GPT-5.4 and GPT-5.4 Pro: What's different

SUMMARY