Model Details
Gemini 3 Flash
Quick answer: Gemini 3 Flash is Google's December 2025 fast multimodal model scoring 78% SWE-Bench Verified, 90.4% GPQA Diamond, 43.5% HLE, 97.5% MATH, and 68.7% SimpleQA. The efficient variant of Gemini 3 — designed for low-latency multimodal tasks.
At a Glance
Where Gemini 3 Flash leads
- 78% SWE-Bench Verified — excellent software engineering for a Flash model
- 90.4% GPQA Diamond — strong graduate-level science
- 97.5% MATH — near-perfect competition math
- 43.5% HLE — competitive frontier reasoning in a Flash tier
- 68.7% SimpleQA — solid factual accuracy
- Multimodal: image + text
Where it lags
- Below Gemini 3.1 Pro (98% ARC-AGI, 94.3% GPQA Diamond) on most metrics
- Proprietary: no open weights
- Flash-class speed/cost tradeoff vs Pro
Best for: High-volume multimodal pipelines requiring strong benchmarks at Flash-tier pricing; software engineering agents; math/science tasks at scale.
What Gemini 3 Flash Is
Gemini 3 Flash is the December 2025 Flash variant in Google's Gemini 3 generation — the faster, cheaper complement to Gemini 3.1 Pro. Like all Gemini Flash models, it is designed for high-throughput deployment with lower latency and cost than the Pro tier, while maintaining strong benchmark performance.
The 78% SWE-Bench Verified is particularly impressive for a Flash-tier model — competitive with GPT-5 Codex (74.5%) and approaching HY3 (78%). Combined with 90.4% GPQA Diamond and 43.5% HLE, Gemini 3 Flash outperforms many models positioned as full-capability systems.
Specifications
| Field | Value |
|---|---|
| Organization | |
| License | Proprietary (API only) |
| Release date | December 2025 |
| Modality | Multimodal (text + image) |
Pricing
Available via Google AI Studio and Google Cloud Vertex AI. Refer to Google AI pricing for Gemini 3 Flash rates.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| SWE-Bench Verified | 78% | Benchgen evaluation | 2025-12 |
| GPQA Diamond | 90.4% | Benchgen evaluation | 2025-12 |
| Humanity's Last Exam | 43.5% | Benchgen evaluation | 2025-12 |
| MATH | 97.5% | Benchgen evaluation | 2025-12 |
| GSM8K | 96.8% | Benchgen evaluation | 2025-12 |
| SimpleQA | 68.7% | Benchgen evaluation | 2025-12 |
Gemini 3 Flash vs Alternatives
| Model | SWE-Bench | GPQA Diamond | HLE | Type |
|---|---|---|---|---|
| Gemini 3 Flash | 78% | 90.4% | 43.5% | Flash |
| Gemini 3.1 Pro | — | 94.3% | 46.4% | Pro |
| HY3 | 78.0% | — | — | General |
| Kimi K2.6 | 80.2% | 90.5% | 36.4% | General |
Gemini 3 Flash nearly matches HY3 on SWE-Bench (78% vs 78%), slightly below Kimi K2.6 (80.2%), and below Gemini 3.1 Pro on GPQA (90.4% vs 94.3%). For Flash-tier multimodal with strong SWE performance: Gemini 3 Flash is the leader.
Related Models & Benchmarks
- Compare: Gemini 3.1 Pro · Kimi K2.6 · HY3
- Benchmarks: SWE-Bench Verified · GPQA Diamond · Humanity's Last Exam
Frequently Asked Questions
What is Gemini 3 Flash?
Google's December 2025 fast multimodal model scoring 78% SWE-Bench Verified, 90.4% GPQA Diamond, 43.5% HLE, 97.5% MATH. Efficient tier of Gemini 3.How does Gemini 3 Flash compare to Gemini 3.1 Pro?
Gemini 3 Flash is faster and cheaper; 3.1 Pro scores higher (94.3% GPQA vs 90.4%, 98% ARC-AGI, 46.4% HLE). Use Flash for high-volume, cost-sensitive pipelines; Pro for maximum capability.Specs from Google's Gemini 3 Flash release (December 2025) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.