Custom ChatGPT: The New ChatGPT-5.1 Is Out. Is It Any Better?

OpenAI shipped ChatGPT-5.1. Naming aside, the release is a meaningful step up across three dimensions: latency, conversational naturalness, and personalization. Below is a practical look at what is new and where it matters in business workflows.

What is genuinely better

1. Latency

First-token latency is materially lower than 5.0, especially on long-context queries. For interactive tools, this matters more than headline benchmark scores.

2. Naturalness

The default voice is less robotic. Less hedging. Less unnecessary scaffolding around responses. For customer-facing applications, including recruiting outreach drafts, this is a real upgrade.

3. Personalization

Custom GPT and Memory features have improved enough to make persistent persona-aware conversations practical. For recruiting specifically, that means agent personas tuned to a specific recruiter's voice are easier to set up.

What is unchanged

• Hallucinations are rarer but still present on long-tail factual queries
• Tool use reliability is improved but not yet at parity with Claude on multi-step plans
• Reasoning under pressure on novel problems is similar to 5.0

ChatGPT-5.1 is not a paradigm shift. It is the kind of solid incremental release that makes a category mature.

How it changes the recruiting use case

For teams using ChatGPT inside their recruiting workflow, the upgrade matters most for outreach drafting and conversational candidate Q&A. The combination of lower latency and better naturalness moves drafts closer to "send as is" quality. Recruiters still review. The friction is just lower.

Vitae customers can bring their own ChatGPT key and route every agent call through the new model with no platform changes required.