OpenAI has introduced another round of updates to ChatGPT, aiming to make its AI assistant faster, more concise, and less likely to present incorrect information with unwarranted confidence. The latest release, centered on the GPT-5.6 family of models, also introduces new controls that allow users to choose how much computational effort ChatGPT spends on answering a prompt.
One of the most notable additions is a new “Think” button, accompanied by a slider that adjusts the amount of processing used to generate responses. Rather than switching between entirely different experiences, users can now balance speed against depth depending on the complexity of their question. Simpler requests can be answered more quickly, while more demanding tasks can draw on additional computing resources to produce more detailed results.
OpenAI says GPT-5.6 has also been trained to avoid several common frustrations users have raised over the past year. Responses are designed to be more direct, reduce unnecessary formatting, and avoid excessive filler language. The company also claims the model is better at challenging incorrect assumptions instead of reflexively agreeing with users, an issue that has drawn criticism as AI assistants become more widely used for research, planning, and everyday problem-solving.
Another headline claim is improved factual reliability. While OpenAI says GPT-5.6 should produce more dependable answers, the company has not suggested that hallucinations have been eliminated. Like other large language models, ChatGPT can still generate inaccurate or fabricated information, meaning users should continue to verify important facts through authoritative sources rather than relying solely on AI-generated responses.
The update also expands access for free users. OpenAI is making GPT-5.6 Luna the default model for free accounts while introducing the new Think option for questions that may benefit from additional processing. The move gives non-paying users access to capabilities that previously required more advanced models or subscription tiers, although usage limits are still expected to apply.
For developers and power users, the compute slider reflects a broader shift in how AI services are being delivered. Rather than offering a one-size-fits-all model, AI providers are increasingly giving users more control over the trade-off between response quality, speed, and infrastructure costs. That flexibility could become more important as demand for AI services continues to grow and computing resources remain expensive.
The latest release also arrives as competition among AI companies intensifies. OpenAI, Google, Anthropic, and other major players are all refining their models to improve accuracy, reduce latency, and make AI assistants feel more useful in day-to-day workflows. Features such as adjustable reasoning modes and better response calibration are becoming key differentiators as companies move beyond raw benchmark performance toward practical usability.
Although GPT-5.6 does not fundamentally change what ChatGPT can do, the update signals OpenAI’s continued focus on refining the user experience. Making responses shorter, reducing misplaced confidence, and giving users greater control over how the model operates may ultimately have a greater impact on everyday use than headline-grabbing increases in model size or benchmark scores.
