Back to AI information
GPT-5 Thinking now features adjustable thinking time: Plus, Pro, and Business users can switch speed and depth on the ChatGPT web app.

GPT-5 Thinking now features adjustable thinking time: Plus, Pro, and Business users can switch speed and depth on the ChatGPT web app.

AI information Admin 149 views

OpenAI has added an adjustable thinking time switch for GPT-5 Thinking: After selecting GPT-5 (Thinking) on the ChatGPT web app, you can switch the thinking time directly in the message input area. Plus and Business offer Standard and Extended; Pro additionally offers Light and Heavy. Your selection will remain in effect in web-based conversations until manually changed.


1. What This Change Addresses

1. Optional Balance between Speed and Intelligence

GPT-5 Thinking now has two new settings: Standard (default, balancing speed and intelligence) and Extended (deeper reasoning). This allows for optimal performance in both daily question-answering and long-term reasoning, reducing the dilemma of being "too slow or too shallow."

2. Pro's four-speed fine control

Pro users have two additional speeds: Light (fastest) and Heavy (deeper), which are suitable for GPT-5 Thinking usage scenarios of different intensities, such as code review, strategy planning, and research writing.

(1) Typical combinations

Light Q&A/drafting: Light or Standard.

Analysis/multi-step planning: Standard or Heavy.

Complex research/long evidence chains: Extended or Heavy.

(2) Key points of the experience

Thinking longer is not necessarily better; GPT-5 can maintain high-quality output with shorter thinking, and it is more cost-effective to switch to a deeper speed when encountering ambiguous tasks.

(3) Persistence settings

In web-based conversations, the gear selection of GPT-5 Thinking will be retained, reducing the cost of repeated switching.


II. Team implementation and cost management

1. Standardized usage

Provide a mapping table of "task→thinking gear" in the team specifications, and put the choice of GPT-5 Thinking in the process to avoid arbitrarily extending the thinking time.

2. Grayscale and observation

Compare the accuracy, rework rate and latency of Standard and Extended in key scenarios, and promote the use of heavy gear after confirming the benefits.

3. Engineering collaboration

(1) Prompt word template

Provide different structured questions for different gears.

(2) Failure fallback

Start with Standard, and switch to Heavy/Extended with one click when necessary.

(3) Billing and monitoring

Include "average time × number of retries × number of words" in the dashboard.

a. Communication Standards

Specify the "expected gear and time limit" in the requirement description.

b. Data Protection

Even if you choose Heavy for sensitive tasks, adhere to the principle of data minimization.

c. Quality Review

Build a library of counterexamples for "when extended thinking time should not be used" to reduce abuse.


Frequently Asked Questions (Q&A)

Q: How do I switch the thinking time in GPT-5 Thinking?

A: Select GPT-5 (Thinking) on the ChatGPT web app. In the message input area, select Light/Standard/Extended/Heavy (depending on your plan) in the Thinking Time switch.

Q: What is the difference between Plus, Pro, and Business gears?

A: Plus and Business offer Standard and Extended; Pro also offers Light and Heavy, covering the full range from extreme speed to deeper reasoning.

Q: Is choosing Extended or Heavy always better?

A: Not necessarily. We recommend starting with Standard to get a baseline, then switching to Extended/Heavy only when stronger reasoning and search integration are needed to control cost and latency.

Q: Will the settings reset every time?

A: No. On the web, the thinking time you choose for GPT-5 Thinking remains in effect until you manually change it.

GPT-5Thinking thinking time can be adjusted GPT-5Thinking gear switching GPT-5 Thinking Standard GPT-5 Thinking Extended GPT-5ThinkingLight GPT-5ThinkingHeavy GPT-5Thinking Speed Intelligent Balance GPT-5Thinking long link reasoning GPT-5Thinking Quick Questions GPT-5Thinking Research Writing GPT-5Thinking Code Review GPT-5Thinking Strategy Planning GPT-5Thinking Default Standard GPT-5Thinking Deeper Reasoning Extended GPT-5Thinking SpeedLight GPT-5Thinking Heavy Reasoning GPT-5Thinking web switch GPT-5Thinking persistence settings GPT-5Thinking conversation remains active GPT-5ThinkingPlus users GPT-5ThinkingPro users GPT-5ThinkingBusiness users GPT-5Thinking Team Specifications GPT-5Thinking Cost Governance GPT-5Thinking grayscale evaluation GPT-5Thinking accuracy comparison GPT-5Thinking rework rate monitoring GPT-5Thinking Delay Control GPT-5Thinking prompt word template GPT-5Thinking Failure Fallback GPT-5Thinking one-click switch GPT-5Thinking Billing Monitoring Average time spent on GPT-5Thinking GPT-5Thinking retries GPT-5Thinking word count overhead GPT-5Thinking Data Minimization GPT-5Thinking sensitive tasks GPT-5Thinking Quality Review GPT-5Thinking counterexample library GPT-5 Thinking When Not Lengthening GPT-5Thinking gear selection guide GPT-5Thinking usage scenarios GPT-5Thinking Q&A recommendation GPT-5Thinking multi-step planning recommendation GPT-5Thinking Complex Research Recommendations GPT-5Thinking retrieval integration GPT-5Thinking Team Mapping Table GPT-5Thinking web page operation GPT-5Thinking settings not resetting GPT-5Thinking usage tips

Recommended Tools

More