Google released Gemini 2.0 Flash Thinking Experimental
Google's first reasoning model, built on the small Gemini 2.0 Flash and released as an experiment in AI Studio and the Gemini API, exposed its thoughts ("shows its thoughts," in Logan Kilpatrick's…
- Date
- 19 December 2024
- Who
- Google DeepMind
- People
- Jeff Dean, Logan Kilpatrick
- Confidence
- Medium
- Deep dive
- Reasoning II, from o1 and o3 to DeepSeek-R1 and the labs that replicated them
Tier: Supporting · Significance: 3/5 · Org(s): Google DeepMind · People: Jeff Dean, Logan Kilpatrick · Confidence: Medium Google's first reasoning model, built on the small Gemini 2.0 Flash and released as an experiment in AI Studio and the Gemini API, exposed its thoughts ("shows its thoughts," in Logan Kilpatrick's phrasing) and was described by Jeff Dean as trained to use thoughts to strengthen reasoning (Forklog report, dated 2024-12-19). It came 98 days after o1-preview and a month before R1, cheaper than o1 with a longer context, yet it drew little market attention; SemiAnalysis later made that point explicitly (SemiAnalysis, "DeepSeek Debates"). Google's reasoning line then restarted at scale with Gemini 2.5.