GPT-5.1 Released: A Warmer ChatGPT With Eight Personality Presets

OpenAI released GPT-5.1 on November 12, 2025, upgrading the GPT-5 series with adaptive reasoning — quick answers for simple questions, deeper thought for hard ones — plus warmer, more conversational responses and eight personality presets for customizing tone.

The changes answered widespread feedback that GPT-5 felt cold: it had turned cautious to cut hallucinations, and 5.1 restored linguistic fluency. OpenAI also shipped GPT-5.1-Codex-Max for long-running coding tasks the same month.

GPT-5.1 signaled a subtle shift: as raw capability saturates, personality and interaction quality become the new competitive axis — the race with Gemini 3 and Claude is no longer just about benchmark scores.

From the GPT-5 Backlash to the 5.1 Correction

The 'warmer' turn was a direct correction of the launch fiasco three months earlier: GPT-5 shipped while force-retiring GPT-4o and other legacy models, and user backlash was fierce enough that OpenAI restored 4o within 24 hours (see our GPT-5 launch coverage). That episode exposed a fact: user attachment to a model is largely bound to tone and personality, not scores. Eight personality presets promote that variable from side effect to product feature; adaptive reasoning is the cost-engineering flip side — simple questions stop burning reasoning tokens, saving compute for genuinely hard ones.

GPT-5.1-Codex-Max, shipped the same month, was the defensive move: with Claude's three-launches-in-three-months run locking up the 'best at coding' mindshare (see our Opus 4.5 coverage), OpenAI had to table an equivalent bid in long-horizon coding.

Our Take

Selling 'warmth' is the tell that consumer AI has entered its retention era: once weekly actives plateau, retention beats acquisition, and personality is retention's first variable. But the path has known minefields — the better personification works, the higher the regulatory stakes around emotional dependency and teen safety (see our Character.AI restriction coverage). After GPT-5.1, 'model personality' is officially a launch-event keyword alongside benchmark scores — a shift that will reshape product design more deeply than any single capability bump.