The wait is finally over GPT-5 is here. After more than two years of anticipation, OpenAI has officially unveiled GPT-5, the successor to GPT-4 and GPT-4o. The new model is already rolling out to all ChatGPT users and introduces a hybrid reasoning engine capable of delivering both rapid responses and in-depth analysis of complex queries.
According to OpenAI, GPT-5 sets a new benchmark in AI performance across a wide range of fields, including coding, mathematics, creative writing, healthcare, and visual interpretation. Unlike traditional models, GPT-5 functions as a comprehensive AI system designed to excel across disciplines.
OpenAI describes GPT-5 as its most advanced coding model to date, with the ability to debug large codebases and generate visually impressive front-end components. Its writing capabilities have also reached new heights able to craft content with emotional depth, rhythm, and a literary touch.
GPT‑5 represents a major advancement in the field of healthcare AI. It significantly outperforms previous models in answering medical queries with clarity and accuracy, making it OpenAI’s most capable model in this domain so far.
When it comes to performance, GPT‑5 raises the bar across multiple industry benchmarks. In the SWE-bench Verified test designed to assess real-world software engineering capabilities GPT-5 scored an impressive 74.9% accuracy when reasoning was enabled. On the Aider Polyglot code editing benchmark, it achieved an 88% pass rate (pass@2) with thoughtful processing.
GPT‑5 Pro also excelled in advanced and highly demanding benchmark evaluations. It secured a notable 42% score with tools in the ultra-challenging Humanity’s Last Exam, and reached an outstanding 89.4% in the GPQA Diamond benchmark, which focuses on PhD-level science questions.
In agentic decision-making benchmarks like Tau2-bench, GPT-5 delivered strong domain-specific results: 96.7% in telecom, 81.1% in retail, and 62.6% in the airline sector. For visual problem-solving, GPT-5 impressed again with an 84.2% score in the MMMU benchmark, showing strong reasoning capabilities.
Perhaps one of the most critical advancements in GPT-5 is its significantly reduced hallucination rate. On the LongFace-Concepts test, it hallucinated only 0.7% of the time an enormous leap forward from the 4.5% recorded by the o3 model.
GPT-5 Availability in ChatGPT
GPT‑5 is now rolling out to free ChatGPT users, starting today, with a limited number of queries available per user. Once that limit is reached, users can continue interacting with the lightweight GPT‑5 Mini model. OpenAI has confirmed that the full reasoning power of GPT‑5 will be gradually made available to all free-tier users over time.
Additionally, GPT‑5 is now the default model for all signed-in ChatGPT users. It officially replaces previous models, including GPT‑4o, OpenAI o3, o4-mini, GPT‑4.1, and GPT‑4.5, marking a significant shift in how users experience the platform.
ChatGPT Plus subscribers ($20/month) get full access to GPT‑5’s advanced reasoning capabilities along with significantly higher usage limits. Meanwhile, ChatGPT Pro users ($200/month) and Team plan users gain access to the more powerful GPT‑5 Pro model, which leverages additional compute and extended reasoning time to tackle complex tasks. OpenAI also confirmed that Enterprise and Edu users will receive access to GPT‑5 within the next week.
For developers, GPT‑5 brings massive improvements to the API, supporting a 400K token context window and outputs of up to 128K tokens. The model is trained on data up to October 1, 2024, ensuring up-to-date performance. When it comes to pricing, access to the GPT‑5 API costs $1.25 per million input tokens and $10 per million output tokens.
