Yesterday, we were all talking about Gemini 3 as the new #1 model.

Today, OpenAI drops two new GPT‑5.1 models: 👉 GPT‑5.1‑Codex‑Max (agentic coding) 👉 GPT‑5.1‑Pro (research‑grade reasoning)

The speed of this industry is just mind-blowing 🤯

1️⃣ GPT‑5.1‑Codex‑Max – frontier coding brain

OpenAI is positioning it as their agentic coding model:

Built on top of the updated reasoning base model
Trained on real software tasks: PR creation, code review, frontend builds, Q&A
Designed for long‑running work: project‑scale refactors, deep debugging, multi‑hour agent loops (this is what interests me most!)
Uses compaction to work coherently across multiple context windows (millions of tokens over a single task)

On benchmarks, it’s now their top coding model: SWE‑bench Verified: 77.9% SWE‑Lancer (IC SWE): 79.9%

For now, it’s only available in Codex surfaces (CLI, IDE extension, cloud, code review), which tells us something interesting: the capabilities aren’t just in the raw model, they’re also codified in the orchestration layer. That’s a fair move to compete with Claude Code, where a lot of the “secret sauce” also lives in orchestration.

2️⃣ GPT‑5.1‑Pro – research‑grade brain 🧠

Alongside that, OpenAI quietly shipped GPT‑5.1‑Pro inside ChatGPT (Pro/Business):

It’s explicitly framed as “research‑grade intelligence”
Much slower, but can think for a long time on hard problems
It’s ChatGPT‑only (no API yet)

I can’t overstate how much I love – and how much I use – the Pro model. I don’t mind waiting if I get a super thorough, genuinely intelligent answer that connects dots all over the place. It’s super impressive.

Very excited to stress‑test GPT‑5.1 Pro.