Google Ships Gemini 2.0 Flash Thinking as a Free Experimental Preview
In one sentence On 19 December 2024 Google released gemini-2.0-flash-thinking-exp, an experimental preview that exposes its own reasoning steps and is free in AI Studio. Google published AIME 2024 at 73.3% and GPQA Diamond at 74.2%, which on AIME is a single question ahead of o1-mini and well behind full o1.
Most AI models answer you straight away. Gemini 2.0 Flash Thinking works differently: before it replies, it writes out its reasoning step by step, and it lets you read that reasoning. Google released it on 19 December 2024 — not as a finished product, but as an experimental preview in AI Studio, free to use within rate limits.
It was Google's first public answer to OpenAI's o1, which had introduced the same idea about three months earlier. On the two hard tests Google chose to publish, the model scored 73.3% on AIME (maths-olympiad problems) and 74.2% on GPQA Diamond (PhD-level science questions). For context, the AIME figure OpenAI publishes for its small reasoning model o1-mini is 70.0% — and AIME has only 30 problems, so that gap is one question. Full o1 was at 83.3%.
The honest summary is therefore not that Google won the reasoning race, but that Google entered it quickly and gave the capability away in a browser tab while the OpenAI equivalent sat behind a paid tier. That pricing asymmetry, more than the scores, is what pushed reasoning models into the mainstream during 2025.
Companies
Google DeepMind
Tools
Gemini 2.0 Flash Thinking
Tags
Sources