Technology & Science
Google Deploys Gemini 3 Flash as Free Default Model Across Search and Gemini App
On 17-18 Dec 2025, Google replaced its 2.5 Flash model with the faster, cheaper Gemini 3 Flash worldwide, making high-level Gemini 3 capabilities the standard for all users and developers at no extra cost.
Focusing Facts
- Gemini 3 Flash became the default in the Gemini app and AI Mode in Google Search on 17–18 Dec 2025, rolling out simultaneously to Google AI Studio, Vertex AI, Antigravity and Android Studio APIs.
- The model processes tasks using about 30 % fewer tokens than Gemini 2.5 Pro and is priced at $0.50 per million input tokens—roughly one-quarter the cost of Gemini 3 Pro.
- Benchmark scores: 81.2 % on MMMU-Pro multimodal reasoning, 90.4 % on GPQA Diamond scientific knowledge, and 33.7 % on Humanity’s Last Exam—near-parity with Gemini 3 Pro.
Context
Tech history shows that once a frontier technology hits a price-performance tipping point, market structure shifts quickly: IBM’s 1981 PC architecture invited cheaper clones that undercut incumbents within five years; Android’s free 2008 release similarly commoditised smartphone software inside a decade. Gemini 3 Flash signals a comparable moment for large-scale AI: Google is collapsing cost and latency while keeping near-flagship accuracy, eroding the premium moat OpenAI and even Google’s own Pro tier relied on. This fits a broader 21st-century trend toward ‘utility AI’ platforms bundled into search, mobile OSs and home devices, accelerating network-effect lock-ins and raising antitrust stakes. Over a 100-year horizon, such moves matter less for any single benchmark score than for how they normalise ubiquitous, low-cost cognitive infrastructure—much as electrification did after 1900—shifting power from standalone model vendors to integrated ecosystem gatekeepers.
Perspectives
Tech and financial publications enthusiastic about Google
finanzen.ch, MoneyControl, PYMNTS, Gizbot, LatestLY — They frame Gemini 3 Flash as a breakthrough that finally combines top-tier reasoning with low latency and price, positioning Google to challenge or even surpass OpenAI. Rely heavily on Google-supplied benchmarks and quotes, so their upbeat coverage risks amplifying corporate marketing while glossing over unanswered questions about real-world accuracy and safety.
South Korean business press
Chosun.com — They treat Gemini’s gains as important but still emphasise that ChatGPT retains overwhelming market share, suggesting the AI landscape is shifting from a single-leader to a two-horse race. By foregrounding local usage statistics and portraying Google mainly as a challenger, the coverage may downplay Gemini’s global momentum to keep the narrative centred on domestic competition.
Consumer-oriented tech blogs
MakeUseOf, 9to5Google — They celebrate Gemini’s arrival on consumer devices and new app integrations as making everyday tasks easier, while noting missing generative features and usage limits. Because these sites build readership around product tips and affiliate links, their reviews tend to spotlight practical perks and only lightly critique strategic or privacy downsides.