AI

Google Launches Gemini 3 Flash — Faster, Smarter Base

Google Gemini 3 Flash: Faster, Smarter Base Model
Gemini 3 Flash
  • Gemini 3 Flash is Google’s new fast base model, replacing the previous Flash as the default in the Gemini app and Search AI Mode.
  • Benchmarks show notable gains: HLE tripled to 33.7% (no tools), Simple QA Verified rose to 68.7%, and coding scores improved significantly.
  • It runs workloads up to three times faster than older Flash models and offers lower per-token costs than Pro for many tasks.
  • Available immediately via the Gemini app, Gemini API, Vertex AI, AI Studio and Antigravity; Pro and image models will remain for heavier workloads.

What is Gemini 3 Flash?

Gemini 3 Flash is the latest member of Google’s Gemini 3 family designed as a faster, more efficient base model. Google positions it as the new default for fast, multimodal responses across consumer and developer surfaces.

Performance and benchmarks

Google reports that Gemini 3 Flash improves accuracy and domain knowledge over the 2.5 Flash lineage. On Humanity’s Last Exam (HLE) the model scores 33.7% without using external tools — roughly three times the previous Flash result and close to Gemini 3 Pro.

Other evaluations are similarly favorable: Simple QA Verified jumps to 68.7% (versus 28.1% for the old Flash), and SWE-Bench Verified shows nearly a 20-point uplift for coding tasks. Google also says Flash matches or beats older models on academic and reasoning tests such as GPQA Diamond and MMMU Pro.

Speed and efficiency

Gemini 3 Flash runs workloads about three times faster than the prior Flash model. Efficiency gains aim to reduce latency for interactive uses like chat, code generation, and multimodal outputs.

Pricing snapshot

Google published per-token pricing for 3 Flash: 1M input tokens = $0.50 and 1M output tokens = $3.00. That is more expensive than 2.5 Flash (input $0.30 / output $2.50) but far cheaper than Gemini 3 Pro (input $2 / output $12) for many use cases.

Availability and where you’ll see it

Gemini 3 Flash is rolling out to the Gemini app and Search’s AI Mode immediately. It becomes the app’s Fast setting and also powers the Thinking option that uses simulated reasoning.

Developers can access 3 Flash today through the Gemini API, Vertex AI, AI Studio and Google’s Antigravity environment. Gemini 3 Pro and the image-focused Nano Banana Pro are expanding in AI Mode as well, but Google has said there will be usage limits for free users.

Why it matters

For most users and many developers, Gemini 3 Flash narrows the gap to Pro-level capabilities while improving speed and lowering costs for common workloads. Organizations that need the highest-end reasoning or image generation can still opt for Gemini 3 Pro or Nano Banana Pro, but Flash should handle a wide range of tasks faster and cheaper.

Read more