OpenAI launches GPT-5, its most advanced model for ChatGPT and API, offering a 256,000-token context window, smarter model routing, and major gains in coding, automation, and scientific reasoning. It outperforms rivals in key benchmarks, adds deeper reasoning controls in the Pro tier, and introduces new pricing tiers, though fine-tuning and continuous learning remain unavailable.
OpenAI has announced the release of GPT-5, its most advanced language model to date, now available to ChatGPT and API users. The new model marks a significant step in the evolution of artificial intelligence for coding, automation, and handling extensive contextual data.
According to OpenAI Chief Executive Officer Sam Altman, GPT-5 represents “a significant step along the path to AGI… a model that is generally intelligent.” The update enhances the core engine that supports millions of AI-driven workflows, setting a higher benchmark for both developers and researchers.
Technical Advancements and Context Expansion
One of the most notable upgrades is the expanded context window of 256,000 tokens. This capability allows the model to process entire codebases, lengthy technical documents, and large datasets without fragmentation.
OpenAI has also implemented smarter model routing, automatically assigning more computational resources to complex problems. The result is improved handling of intricate multi-step tasks, tool usage, and instruction following — key to agentic automation.
Performance Across Benchmarks
GPT-5 surpasses previous OpenAI models and many industry competitors in several areas:
- Coding performance: Achieves 74.9% on SWE-bench Verified, outperforming Claude Opus 4.1 (74.5%) and Gemini 2.5 Pro (59.6%).
- Multilingual code editing: Reaches 88% on the Aider Polyglot benchmark.
- Scientific reasoning: The GPT-5 Pro variant scores 89.4% on the GPQA Diamond benchmark, representing PhD-level performance.
- Long-context processing: Capable of working with up to 256,000 tokens in a single task.
- Automation: Manages complex agentic tasks with improved instruction following and tool integration.
New and Enhanced Features
Compared to earlier models, GPT-5 delivers several key improvements:
- Context window expansion far beyond previous limits.
- Automated model selection for each query, optimising computational efficiency.
- Improved code reliability and error handling, reducing developer friction.
- “GPT-5 Thinking” in the Pro tier, offering greater reasoning depth and control.
- Adjustable verbosity and output length through the API.
Where GPT-5 Is Strong but Not Market-Leading
While the model achieves substantial gains, OpenAI notes that some areas remain competitive rather than dominant. In certain agentic benchmarks, GPT-5 delivers performance similar to top rivals rather than exceeding them. Everyday writing quality shows moderate improvement, and although output remains consistent, verification is still recommended for high-stakes applications.
Limitations and Features Yet to Arrive
Some capabilities remain unchanged or absent from this release:
- Public fine-tuning is not available at launch.
- The model does not update after deployment, meaning no continuous learning.
- Automated routing may occasionally misjudge the complexity of a task without clear user guidance.
Pricing and Migration Guidance
OpenAI has outlined pricing across multiple variants:
- gpt-5: $1.25 per million input tokens, $10 per million output tokens.
- gpt-5-mini: $0.25 per million input tokens, $2 per million output tokens.
- gpt-5-nano: $0.05 per million input tokens, $0.40 per million output tokens.
- Pro tier: $200 per month, offering unlimited usage and enhanced reasoning depth.
For migration, OpenAI advises users to update the model name in the API and review workflows to assess any changes in outputs.
OpenAI’s GPT-5 release underscores the organisation’s focus on advancing computational reasoning, large-scale context handling, and automation, while leaving scope for further refinement in real-time adaptability and customisation.
