OpenAI recently released upgrades to their GPT-5 model. GPT‑5.1 Instant, the default chat model, has improvements to instruction following. GPT‑5.1 Thinking, the reasoning model, is faster and gives more understandable responses. GPT‑5.1-Codex-Max, the coding model, is trained to use compaction to perform long-running tasks.
The Instant and Thinking models are used in the ChatGPT application, and are chosen automatically by an auto-routing model. OpenAI says that users wanted models that are enjoyable to talk to, leading the company to change their models’ default response style. The new release also includes more customization control over chat tone and style. ChatGPT already had several preset styles, and the release includes three new styles: Quirky, Candid, and Professional. Users can also tweak response style with personalization settings and custom instructions.
OpenAI’s rollout of GPT-5 earlier this year ignited a controversy when they removed the option to choose older models, breaking many user workflows. While the company continues to have ChatGPT automatically choose the best model for a chat, OpenAI reversed its plan to immediately deprecate older models. According to OpenAI:
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.
The updates to the model’s response settings were part of OpenAI’s effort to make it easier for users to make ChatGPT’s personality “[feel] uniquely right.” The company says that the preset response styles were developed based on “what we’ve learned about how people naturally steer the model.” While the presets were rolled out to everyone with the release, the fine-tuning settings are being released to a few users at a time, as an experiment.
Several Hacker News users were not impressed with the conversational tone of the model, expressing a preference for more terse responses. One user pointed out that the “Efficient” (formerly “Robot”) preset was intended for this. Another wrote:
Seems like people here are pretty negative towards a “conversational” AI chatbot. ChatGPT has a lot of frustrations and ethical concerns, and I hate the sycophancy as much as everyone else, but I don’t consider being conversational to be a bad thing. It’s just preference I guess.
GPT‑5.1-Codex-Max is an improved version of OpenAI’s GPT‑5.1-Codex coding model. In tests on the SWE-bench Verified benchmark, GPT‑5.1-Codex-Max outperformed the previous model while using fewer reasoning tokens. The model also has an Extra High (xhigh) reasoning effort level, in addition to the previous low, medium, and high levels.
GPT-5.1-Codex-Max Benchmark Performance. Image Source: OpenAI Blog
Hacker News users also discussed GPT‑5.1-Codex-Max:
Codex CLI is great and I use it often but I’d like to have more of these convenient features for managing context from CC. I’m super happy that compaction is now available, hopefully we’ll get more features for managing context.
Another user said:
It would be great to have access to this model via the chat interface, even if it was gated behind the “other models” dropdown or something.
The new Codex model is available in the Codex CLI and several IDE extensions. API access is “coming soon.”
