GPT 4.1 just dropped!

Limited info so here's what I know:

• multimodal
• excels at coding
• 1m context
• possibly the cloaked models we saw earlier in the week (huge if true)

Pricing is competitive with other models in its class.

Claude has a competitor to Deep Research

It's called "Research"
Currently only available on the max plan, because of course it is

o3 & o4 mini!

Here's what's known:

• o3 pulled 20.3 which is a new record for Humanity's Last Exam (kaileh.dev/hle)
• o3 & o4 mini are trading blows with each other and Gemini 2.5 pro across the benchmark suite
• Google's models are no longer the only ones competent at math, and actually lose sometimes to the OpenAI ones
• My best guess is that this is o3-medium, as there's no way they got the performance cost of o3-high from the ARC AGI prize down this low this quickly
• They also released Claude code but with ChatGPT

Hey look it's Gemini 2.5 flash!

• Hybrid reasoning and direct answer model
• Overall very solid uplift over 2.0 flash, keep in mind this is a model that responds nearly instantly