GPT-5.1 Launch: Instant, Thinking, and Eight Personalities

Two variants, eight personalities

OpenAI released GPT-5.1 on November 12, 2025 as two variants: GPT-5.1 Instant, described as ChatGPT's most-used model made "warmer by default and more conversational," and GPT-5.1 Thinking, aimed at harder problems. Both models now vary how long they spend reasoning based on the apparent difficulty of the question, running faster on simple tasks and slower on ones that need more work, and OpenAI said Thinking's answers came with clearer language and fewer undefined technical terms. [1][2]

Alongside the model change, OpenAI expanded ChatGPT's personality presets to eight: Default, Professional, Friendly, Candid, Quirky, Efficient, Nerdy, and Cynical. Two of the six that existed before the update were renamed as part of the expansion, with Friendly replacing a preset called Listener and Efficient replacing one called Robot. Each preset changes how concise, warm, or emoji-heavy responses are, not the underlying model. [2][3]

Atlas interpretation: The summary's claim that tone of voice was the headline feature of a frontier release is borne out by how OpenAI itself sequenced the announcement: variant names and adaptive reasoning came first, but the eight-preset personality system and the warmth of the Instant model's default voice got equal or greater billing. A model update was being sold partly as a mood update. [1][2]

The tone problem it was answering

GPT-5.1 arrived roughly three months after GPT-5's launch, which drew user complaints that the new model felt colder and more mechanical than GPT-4o even where it scored better on technical benchmarks. Coverage of the GPT-5.1 launch read the new personality presets and the warmer Instant default as OpenAI's follow-through on that complaint rather than a standalone feature. [2][4]

Atlas interpretation: Independent commentary split on whether that follow-through was an improvement. Coverage from Stark Insider treated the warmer default and expanded presets as OpenAI directly answering complaints about GPT-5's coldness. John Gruber, writing on Daring Fireball the same week, took the opposite view, calling phrases like "I've got you" a "glaring regression" and arguing OpenAI had reversed an earlier, more defensible instinct toward a plainer default tone because enough users wanted the warmth back. Both readings agree on the underlying fact: the new default voice was a deliberate response to user preference, not a byproduct of a capability upgrade. [3][4]

Gruber's own preference landed on the Efficient preset, the renamed successor to Robot, for stripping out the emotional language and emoji he objected to in the new default. OpenAI's release notes describe the change more neutrally, as making "GPT-5's default personality warmer and more familiar." [3][5]

A staged rollout, old model kept on standby

Paid Plus, Pro, Go, and Business subscribers got access first, with free and logged-out users following and Enterprise and Edu users given a seven-day toggle before GPT-5.1 became their default. OpenAI kept the prior GPT-5 model available to paid subscribers for three months after the switch. [2]

Atlas interpretation: That three-month overlap is the same lesson OpenAI drew from the GPT-5 rollout, when pulling GPT-4o access without notice produced the loudest part of the backlash. Keeping the outgoing model reachable this time turned a repeat of that specific complaint into a decision users could opt out of instead of one imposed on them. [4]