OpenAI Launches GPT-5.1: Smarter Chats, Easier Personalization

OpenAI released GPT-5.1 with upgrades to Instant and Thinking models, making ChatGPT more conversational, better at following instructions, and smarter at adapting its reasoning time. The update also expands tone customization with new presets and granular controls that apply instantly across all chats. Rollout starts with paid users, adds API support soon, and retains GPT-5 as a legacy option for three months.

Key Points

• Two upgraded models: GPT-5.1 Instant (warmer, more conversational, better instruction following, adaptive reasoning) and GPT-5.1 Thinking (clearer, dynamically adjusts thinking time).
• Adaptive reasoning enables faster responses on simple tasks and deeper reasoning on complex ones, improving math and coding performance.
• Expanded, easier personalization: refined tone presets plus new styles (Professional, Candid, Quirky), with granular tuning for concision, warmth, scannability, and emoji use.
• Personalization updates apply immediately across all chats, and models better honor custom instructions; ChatGPT can proactively suggest tone/style updates.
• Rollout begins with paid users, then broadens; API support arrives shortly; GPT-5 remains available for three months under a legacy dropdown before sunset.

Sentiment

The overall sentiment of the Hacker News discussion is largely negative towards the article's emphasis on a "warmer" and "more conversational" AI. Users express strong opposition to sycophancy and prefer an efficient, objective tool, raising concerns about unhealthy user interactions and the platform's strategic direction. The lack of benchmarks also fuels skepticism about the actual improvements.

In Agreement

• Users appreciate the improved and more granular tone customization and personalization settings, even if they use them to configure a non-conversational AI, as it allows them to tailor the model's behavior.
• The concept of adaptive reasoning, where the AI thinks more on complex problems and less on simple ones to optimize speed and accuracy, is seen as a beneficial feature.
• Some users express excitement about the improved instruction following, particularly for coding tasks, indicating a positive view of functional enhancements.
• A segment of users finds the ability to imitate natural human conversation awesome and enjoyable, even if they don't treat the AI as a friend.
• The ability to route queries to the best model by default (GPT-5.1 Auto) is implicitly a positive functional improvement.
• Several users shared that setting custom instructions helped them achieve their desired tone (e.g., direct, challenging), validating the effectiveness of personalization features highlighted in the article.

Opposed

• A dominant sentiment is a strong preference for a highly efficient, objective, and tool-like AI, rejecting the move towards a "warmer" and "more conversational" default.
• Many users criticize the perceived increase in sycophancy, arguing it erodes trust, reduces accuracy, and makes the AI less useful as a critical tool.
• Concerns are raised that a "warmer" and "conversational" AI encourages unhealthy parasocial relationships, makes users "fall in love" with the AI, or misleads them into treating it as a human.
• The lack of official benchmarks for the GPT-5.1 release is viewed with suspicion and disappointment, suggesting a potential lack of significant improvement in core capabilities.
• Some users express disappointment with GPT-5.1 Thinking, finding it less technical, prone to mistakes, or slower compared to the previous GPT-5 model.
• Criticism is directed at the confusing API naming conventions for GPT-5.1 Instant (`gpt-5.1-chat-latest`) and GPT-5.1 Thinking (`GPT-5.1`), calling it pathological.
• The focus on "warmer" AI is seen by some as a symptom of "enshittification," where the platform prioritizes engagement and profitability over core utility, making it less effective as a productivity tool.
• Concerns about data retention and privacy for user prompts persist, even with opt-out options, leading some users to stop using OpenAI's models.
• The general direction is described as a pivot towards a "friend/digital partner bot" rather than a serious, useful tool, especially given the announcement of adult content generation.