For once it is Friday as I write this, everyone should be very proud. Anyway, here's the stories you were definitely desperate for.

Google had a busy week! Several new models came out and, to be honest, Google's back in the AI race!

2.0 Flash: This small version of Gemini is incredibly quick and quite a bit cheaper than the others. It's free to access on gemini.google.com. The extreme speed and cost more than makes up for the only decent performance, but it's definitely a contender with 4o in several areas, and is especially good at math.

2.0 Pro: The bigger and slower cousin of flash, it does a bit better at benchmarks at the cost of higher prices and slower response times.

2.0 Flash Thinking (with apps): Google has their first real chain-of-thought (thinking) model. Like DeepSeek R1 and OpenAI o3, this model thinks through its steps and plans before responding. Gemini's performance is pretty good, and the ability to interact with apps (such as YouTube, Maps, Messages, Calendar etc) makes it a genuinely useful assistant!

As a reminder, Google's main strength so far has been speed, price, and context. Context is the amount of data that a model can use when creating a response, and in the case of Gemini it's over 20x their competitors at OpenAI. This means it can deal with over 5000 pages of text, dozens of hours of audio, or up to 2 hours of video and keep all of it in mind when responding to you!

Guess it's Gemini week... Yeah there doesn't seem to be any activity from other teams... Still good to see Google back in the race