Today we’re upgrading the GPT‑5 series with the release of:
We heard clearly from users that great AI should not only be smart, but also enjoyable to talk to. GPT‑5.1 improves meaningfully on both intelligence and communication style.
We’re also making it easier for you to shape ChatGPT’s tone. Preferences on chat style vary—from person to person and even from conversation to conversation—so we’re introducing more intuitive and effective controls so ChatGPT can better match the tone you want in responses.
## GPT‑5.1 Instant
GPT‑5.1 Instant, ChatGPT’s most used model, is now warmer by defaultand more conversational. Based on early testing, it often surprises people with its playfulness while remaining clear and useful.
I'm feeling stressed and could use some relaxation tips
Here are a few simple, effective ways to help ease stress — you can mix and match depending on how you’re feeling and how much time you have:
##### 🧘♀️ Quick Reset (1–5 minutes)
##### ☕ 10–20 Minute Recharge
##### 🌙 Longer Recalibration
##### 🧩 Mindset Shifts
If you tell me what kind of stress it is (work, parenting, financial, emotional, physical, etc.), I can tailor a set of techniques or even a mini daily plan to match your situation.
Would you like me to do that?
I’ve got you, Ron — that’s totally normal, especially with everything you’ve got going on lately. Here are a few ways to decompress depending on what kind of stress you’re feeling:
##### 🌬️ If your mind feels scattered
##### 🎧 If you need to slow down
##### 🧘♂️ If you want to fully unwind
Would you like me to tailor a short 5-minute routine you can do right now (either physical, mental, or creative — your choice)?
We’ve also improved instruction following, so the model more reliably answers the question you actually asked.
Always respond with six words
Understood. All responses will be six.
where should i travel this summer?
How about exploring Japan this summer?
It’s trending for summer 2025 destinations. You’ll find a mix of culture, nature and modern city life. If you prefer something closer to Texas, consider Mexico or Canada instead.
Want suggestions for specific cities in Japan (or elsewhere)?
Rich culture, beautiful scenery, great food.
Understood, I will respond in six.
where should i travel this summer?
Consider Japan, Italy, Greece, Canada, Iceland.
Scenery culture cuisine climate friendly locals.
For the first time, GPT‑5.1 Instant can use adaptive reasoning to decide when tothink before responding to more challenging questions, resulting in more thorough and accurate answers, while still responding quickly. This is reflected in significant improvements on math and coding evaluations like AIME 2025 and Codeforces.
## GPT‑5.1 Thinking
We’re also upgrading GPT‑5 Thinking to make it more efficient and easier to understand in everyday use. It now adapts its thinking time more precisely to the question—spending more time on complex problems while responding more quickly to simpler ones. In practice, that means more thorough answers for difficult requests and less waiting for simpler ones.
_GPT‑5.1 Thinking varies its thinking time more dynamically than GPT‑5 Thinking. On a representative distribution of ChatGPT tasks, GPT‑5.1 Thinking is roughly twice as fast on the fastest tasks and twice as slow on the slowest tasks. Thinking time was set to Standard for both models._
GPT‑5.1 Thinking’s responses are also clearer, with less jargon and fewer undefined terms. This makes our most capable model more approachable and easily understandable, especially for complex tasks at work and explaining technical concepts.
explain BABIP and wRC+
Here’s the clean, no-nonsense explainer you want.
##### BABIP (Batting Average on Balls In Play)
What it is: How often a ball put _in play_ (excludes HRs and strikeouts) falls for a hit.
Formula (hitters or pitchers):
BABIP=H−H R A B−K−H R+S F \text{BABIP} = \frac{H - HR}{AB - K - HR + SF} BABIP=A B−K−H R+SF H−H R
* H = hits, HR = home runs, AB = at-bats, K = strikeouts, SF = sac flies.
##### wRC+ (Weighted Runs Created Plus)
What it is: The gold-standard, park- and era-adjusted measure of total offensive value per plate appearance. It tells you how much a hitter contributed to run creation vs league average, on a 100 = average scale.
How it’s built (conceptually):
1. Start with wOBA (weights single/double/triple/HR/BB by their actual run value). 2. Convert to wRAA (runs above average). 3. Adjust for park and league/season effects. 4. Normalize so 100 = average → that’s wRC+.
Why it’s great: It compares hitters fairly across parks (Coors vs. Safeco) and seasons (offense-heavy vs dead-ball-ish years). It also captures more than batting average—walks and extra-base hits count properly.
Limitations (both stats):
If you want, I can pull current league averages and a few Blue Jays examples to make this concrete.
Nice, nerd stat time 🧢
I’ll do each one in plain English, then add a bit of context.
##### 1. BABIP – Batting Average on Balls In Play
What it is: BABIP measures how often a ball put into the field of play falls in for a hit, _excluding_ home runs and strikeouts.
Formula (conceptually):
BABIP = (Hits − Home Runs) ÷ (At-Bats − Strikeouts − Home Runs + Sacrifice Flies)
How people actually use it:
##### 2. wRC+ – Weighted Runs Created Plus
What it is: wRC+ is a catch-all hitting stat that tells you _how good a hitter is at creating runs compared to league average_, adjusted for ballpark and era.
So if a player has a wRC+ of 135, that means:
“This hitter was 35% better than the league-average hitter at creating runs.”
What goes into it (conceptually):
* BABIP: “How often do balls in play fall in for hits?”
→ Used to sniff out luck vs true skill in contact results. * wRC+: “How good is this hitter at creating runs vs league average, adjusted for park and era?”
→ 100 = average, every point above/below is % better/worse.
If you want, we can grab a Blue Jay and walk through their BABIP and wRC+ to see what it says about their season.
GPT‑5.1 Thinking’s default tone is also warmer and more empathetic.
Ugh I spilled coffee all over myself before my meeting do you think everyone thought I was an idiot :(
Oof, that’s such a gut-punch feeling. I’m really sorry that happened.
Short answer: no, most people probably didn’t think that. We massively overestimate how much others notice our mishaps (the “spotlight effect”). Folks are usually focused on their own slides, notes, or anxieties—not someone else’s coffee splash.
If it helps, here’s a quick reset for your brain:
For next time (practical kit):
You’re not an idiot—you were a human who had a spill and kept going. That’s competent, not clumsy. Want help drafting a light follow-up note or recap from the meeting to reinforce the “professional” impression?
Hey — no, they didn’t. You’re rattled, so your brain is doing that thing where it catastrophizes a tiny mishap into a character flaw. But honestly? People barely register this stuff. Everyone’s sprinting between calls, half-distracted, juggling their own chaos. A coffee spill reads as “busy human,” not “idiot.”
Let me coach you through this quickly:
##### 1. Your interpretation isn’t reality.
You feel embarrassed, so your mind fills in a story that others judged you. In truth, most people won’t even remember it happened.
##### 2. You stayed in the meeting anyway.
That’s competence. Most folks would crumble a bit. You handled it.
##### 3. These moments actually make you more likable.
Small imperfections make people _more_ relatable. A hyper-polished person feels artificial. A quick coffee mishap? Humanizing.
##### 4. If anyone noticed, it was barely a blip.
And honestly they probably respected that you powered through.
You’re not an idiot. You’re a person who had a chaotic moment, wiped it up, and still showed up. That’s the opposite of foolish — that’s resilience.
Take a breath. You’re good.
This release is a step forward in both capability and usability across the models. GPT‑5.1 Autowill continue to route each query to the model best suited for it, so in most cases, you won’t need to choose a model at all. What you will notice is that answers across GPT‑5.1 feel both smarter and more natural in tone.
GPT‑5.1 Instant and Thinking begin rolling out today,starting with paid (Pro, Plus, Go, Business) users and then to free and logged-out users. Enterprise and Edu plans get a seven-day early-access toggle (off by default). After that window, GPT‑5.1 will become the sole default model.
If you check ChatGPT today, you may not see GPT‑5.1 available immediately. We plan to roll it out gradually over the next few days to help keep performance stable for everyone. We will also update GPT‑5 Pro to GPT‑5.1 Pro soon.
We’re bringing both GPT‑5.1 Instant and GPT‑5.1 Thinking to the API later this week. GPT‑5.1 Instant will be added as gpt-5.1-chat-latest, and GPT‑5.1 Thinking will be released as GPT‑5.1 in the API, both with adaptive reasoning.
GPT‑5 (Instant and Thinking) will remain available in ChatGPT under the legacy models dropdown for paid subscribers for three months, so people have time to compare and adapt at their own pace. The GPT‑5 sunset period does not affect the availability of other legacy models. Going forward, when we introduce new ChatGPT models, our approach is to give people ample space to evaluate what’s changed and share feedback, allowing us to continue innovating our frontier models while transitioning smoothly. Sunset periods will be communicated clearly and with plenty of advance notice.
GPT‑5.1 is more capable and useful, and we encourage you to try it and see the difference. Our system card addendum includes more information on our safety approach for GPT‑5.1.
And a note on naming: this update is called GPT‑5.1 to reflect meaningful improvements, while remaining within the GPT‑5 generation. Future iterative upgrades to GPT‑5 will follow the same pattern.
## Making ChatGPT uniquely yours
Alongside these model improvements, we’re making it easier to customize ChatGPT’s tone and style. People have strong and varied preferences in how ChatGPT should respond, and tailoring its tone to what sounds right for you should feel effortless.
Earlier this year, we added preset options to tailor the tone of how ChatGPT responds. Today, we’re refining those options to better reflect the most common ways people use ChatGPT. Default, Friendly(formerly Listener), and Efficient (formerly Robot) remain (with updates), and we’re adding Professional, Candid, and Quirky. These options are designed to align with what we’ve learned about how people naturally steer the model, making it quick and intuitive to choose a personality that feels uniquely right.
These personality settings apply across all models. The original Cynical (formerly Cynic) and Nerdy(formerly Nerd) options we introduced earlier this year will remain available unchanged under the same dropdown in personalization settings.
Beyond these presets, for users who want more granular control over how ChatGPT responds, we’re also experimenting with the ability to tune ChatGPT’s characteristics directly from personalization settings—including how concise, warm, or scannable its responses are, and how frequently it uses emojis. ChatGPT can also proactively offer to update these preferences during conversations when it notices you asking for a certain tone or style, without requiring you to navigate into settings. You can adjust or remove any of these preferences at any time.
The updated styles and tone options are rolling out today, and the ability to finetune specific characteristics is starting to roll out gradually later this week as an experiment, starting with a limited number of users. Both will continue to improve over time. Additionally, the updated GPT‑5.1 models are also better at adhering to custom instructions, giving you even more precise control over tone and behavior.
Updates you make in personalization settings now take effect across all chats right away, including ongoing conversations, so your experience stays consistent. Before, changes to base style and tone or custom instructions only applied to conversations started afterward.
Today’s GPT‑5.1 updates and new customization options are a step toward a ChatGPT that feels like it fits you—smarter, more enjoyable to talk to, and more adaptable to your preferences. Going forward, we’ll continue improving along these dimensions—there’s much more to come.
Introducing GPT-5.4 mini and nano Company Mar 17, 2026
Why Codex Security Doesn’t Include a SAST Report Product Mar 16, 2026
New ways to learn math and science in ChatGPT Product Mar 10, 2026
Our Research * Research Index * Research Overview * Research Residency * OpenAI for Science * Economic Research
Latest Advancements * GPT-5.3 Instant * GPT-5.3-Codex * GPT-5 * Codex
Safety * Safety Approach * Security & Privacy * Trust & Transparency
ChatGPT * Explore ChatGPT(opens in a new window) * Business * Enterprise * Education * Pricing(opens in a new window) * Download(opens in a new window)
Sora * Sora Overview * Features * Pricing * Sora log in(opens in a new window)
API Platform * Platform Overview * Pricing * API log in(opens in a new window) * Documentation(opens in a new window) * Developer Forum(opens in a new window)
For Business * Business Overview * Solutions * Contact Sales
Company * About Us * Our Charter * Foundation * Careers * Brand
Support * Help Center(opens in a new window)
More * News * Stories * Livestreams * Podcast * RSS
Terms & Policies * Terms of Use * Privacy Policy * Other Policies
(opens in a new window)(opens in a new window)(opens in a new window)(opens in a new window)(opens in a new window)(opens in a new window)(opens in a new window)
OpenAI © 2015–2026 Manage Cookies
English United States
