OpenAI ships GPT-5.1 with adaptive reasoning and warmer tone

OpenAI began rolling out GPT-5.1 to ChatGPT on Tuesday, splitting the upgrade into two variants — Instant and Thinking — that introduce adaptive reasoning to the company's most-used model for the first time.

The release arrives two months after GPT-5 debuted in August. OpenAI labeled it "GPT-5.1" rather than GPT-6 to flag "meaningful improvements, while remaining within the GPT-5 generation." Future iterative updates will use the same pattern, the company said.

What changed in GPT-5.1 Instant?

GPT-5.1 Instant, the default for most ChatGPT queries, now arrives "warmer by default and more conversational," OpenAI said. It also picked up adaptive reasoning — the ability to pause and think before answering harder prompts, something previously reserved for Thinking. On math and coding tests including AIME 2025 and Codeforces, the company reported "significant improvements."

Three shifts sit inside the Instant upgrade:

• Warmer, more conversational default tone
• Better instruction following — the model answers what was asked
• Adaptive reasoning for difficult questions, with fast replies on easy ones

How does Thinking change?

GPT-5.1 Thinking — formerly GPT-5 Thinking — varies its thinking time more dynamically. 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, OpenAI said. Thinking time was set to Standard for both models in the comparison.

The variant's responses also dropped jargon. "Less jargon and fewer undefined terms" make the model "more approachable and easily understandable," according to OpenAI. The default tone is warmer and more empathetic.

GPT-5.1 Auto, the router that picks a model per query, remains in place. Most users will not need to choose.

Who gets access first?

The rollout tier order:

• Today: Pro, Plus, Go, and Business subscribers
• Following days: free and logged-out users
• Enterprise and Edu: seven-day early-access toggle (off by default)
• After the toggle window: GPT-5.1 becomes the sole default

GPT-5 Pro will update to GPT-5.1 Pro "soon." GPT-5 (Instant and Thinking) stays in the legacy models dropdown for paid subscribers for three months. The sunset period does not affect availability of other legacy models.

The API follows later this week. GPT-5.1 Instant ships as gpt-5.1-chat-latest. GPT-5.1 Thinking ships as GPT-5.1. Both expose adaptive reasoning to developers.

Why does the personality rework matter?

OpenAI overhauled ChatGPT's tone presets. The earlier Default, Listener, and Robot gave way to Default, Friendly (the renamed Listener), and Efficient (the renamed Robot). Three new options joined the menu:

• Professional — polished and precise
• Candid — direct and encouraging
• Quirky — playful and imaginative

The original Cynical and Nerdy presets remain unchanged under the same personalization dropdown.

These personality settings apply across every model. OpenAI is also testing finer-grained sliders in personalization settings — controls for how concise, warm, or scannable a reply is, plus a knob for emoji frequency. ChatGPT can proactively offer to update those preferences mid-conversation when it detects the user asking for a certain tone. Users can adjust or remove any preference at any time.

Personalization changes now propagate to every chat immediately, including ongoing threads. Previously, changes only applied to conversations started after the edit. GPT-5.1 models also adhere more reliably to custom instructions, the company said.

What does this mean for ChatGPT users?

For most users, the Instant upgrade delivers two visible shifts: shorter waits on easy prompts and more substantive answers on hard ones, without a manual model switch. The Thinking upgrade redistributes latency — fast tasks get faster, hard tasks take longer — and trades jargon for clearer prose.

OpenAI framed the dual release as a response to feedback. "We heard clearly from users that great AI should not only be smart, but also enjoyable to talk to," the company wrote. "GPT-5.1 improves meaningfully on both intelligence and communication style."

The customization push points to a broader product direction. OpenAI is positioning ChatGPT as a configurable assistant whose tone matches the user rather than a single voice for everyone. Granular sliders and proactive preference updates suggest tone is becoming a first-class setting, on par with model selection.

What's next?

OpenAI published a system card addendum covering GPT-5.1's safety approach. The company did not specify when GPT-5.1 will fully replace GPT-5 as default beyond the seven-day Enterprise and Edu toggle. GPT-5 lingers in the legacy dropdown for three months as a safety net.

The release also sharpens a competitive question. Anthropic, Google, and a growing roster of open-weight labs have shipped conversational tuning of their own in recent months. OpenAI's bet — a warm, customizable default plus adaptive reasoning across both Instant and Thinking tiers — sets a target competitors will now have to match or beat, not just on benchmark scores but on the harder problem of how an assistant should sound.