Google model product
Gemini 3 Flash offers frontier intelligence built for speed at a fraction of the cost.
Updated Aug 12, 2026. Default version: Gemini 3 Flash
Technical details for the model's default version.
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Benchmark scores for Gemini 3 Flash.
| Global PIQA | 92.8% | 02 | 13 | 91.7% | C | |
| t2-bench | 90.2% | 02 | 23 | 95.5% | C | |
| VideoMMMU | 86.9% | 02 | 26 | 96.0% | C | |
| MMMLU | 91.8% | 03 | 49 | 95.8% | C | |
| LiveCodeBench Pro | 2,316 points | 04 | 4 | 0.0% | C | |
| Vending-Bench 2 | 3,635 usd | 04 | 4 | 0.0% | C | |
| AIME 2025 | 99.7% | 07 | 114 | 94.7% | C | |
| SimpleQA | 68.7% | 07 | 46 | 86.7% | C | |
| MMMU-Pro | 81.2% | 08 | 66 | 89.2% | C | |
| FACTS Grounding | 61.9% | 10 | 13 | 25.0% | C | |
| Legal Agent Benchmark | 0.0% | 10 | 13 | 25.0% | B | |
| ScreenSpot Pro | 69.1% | 10 | 25 | 62.5% | C | |
| ARC-AGI v2 | 33.6% | 11 | 16 | 33.3% | C | |
| Finance Agent v2 | 42.5% | 14 | 26 | 48.0% | B | |
| Toolathlon | 49.4% | 15 | 31 | 53.3% | C | |
| OmniDocBench 1.5 | 12.1% | 16 | 17 | 6.3% | C | |
| MRCR v2 (8-needle) | 22.1% | 17 | 21 | 20.0% | C | |
| SWE-Bench Verified | 78.0% | 19 | 105 | 82.7% | C | |
| CharXiv-R | 80.3% | 20 | 48 | 59.6% | C | |
| GPQA | 90.4% | 20 | 234 | 91.8% | C | |
| LiveBench | 72.4% | 24 | 38 | 37.8% | B | |
| Humanity's Last Exam | 43.5% | 29 | 93 | 69.6% | C | |
| MCP Atlas | 57.4% | 29 | 31 | 6.7% | C | |
| Terminal-Bench 2.0 | 47.6% | 37 | 49 | 25.0% | C |
Preference and agent-evaluation results for the default version.
| search | overall | 13 | 1197.3 | 125,928 | N/A | |
| search factuality | overall | 17 | 1180.2 | 125,762 | N/A | |
| search style control | overall | 17 | 1190.2 | 125,928 | N/A | |
| text | overall | 24 | 1465.8 | 30,861 | N/A | |
| text factuality | overall | 24 | 1466.8 | 30,248 | N/A | |
| vision | overall | 24 | 1283.4 | 35,759 | N/A | |
| text style control | overall | 28 | 1472.8 | 30,861 | N/A | |
| vision style control | overall | 28 | 1270.8 | 35,759 | N/A | |
| document style control | overall | 29 | 1440.8 | 7,173 | N/A | |
| vision style control | overall | 33 | 1259.0 | 34,207 | N/A | |
| agent steerability | overall | 34 | -0.0 | N/A | 85.8K | |
| document | overall | 34 | 1413.3 | 7,173 | N/A | |
| vision | overall | 35 | 1264.6 | 34,207 | N/A | |
| agent tool hallucination | overall | 37 | -0.0 | N/A | 1.6M | |
| agent task outcome explicit | overall | 38 | -0.1 | N/A | 59.1K | |
| agent praise complaint | overall | 40 | -0.1 | N/A | 21.5K | |
| agent | overall | 41 | -0.1 | N/A | 1.8M | |
| agent bash recovery steps | overall | 44 | -0.2 | N/A | 78.7K | |
| webdev | overall | 46 | 1438.1 | 10,481 | N/A | |
| text style control | overall | 47 | 1458.2 | 86,493 | N/A | |
| text factuality | overall | 50 | 1451.4 | 85,736 | N/A | |
| text | overall | 54 | 1442.2 | 86,493 | N/A | |
| webdev | overall | 70 | 1383.1 | 20,932 | N/A |
Provider-specific output speed and catalog latency for Gemini 3 Flash. Runtime does not affect the capability score.
| 61.686 tok/s | 6.001 s | 1M | 65.5K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Official vendor API pricing appears first, followed by individual provider offers.
| QiHang | gemini-3-flash-preview | global | $0.07 | $0.43 | 1M | |
| Poe | google/gemini-3-flash | global | $0.40 | $2.4 | 1M | |
| gemini-3-flash-preview | global | $0.50 | $3 | 1M | ||
| NanoGPT | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| LLM Gateway | gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| OpenCode Zen | gemini-3-flash | global | $0.50 | $3 | 1M | |
| 302.AI | gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| FrogBot | gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| ZenMux | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Abacus | gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| AIHubMix | gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Ofox | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| OrcaRouter | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Vertex | gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Kilo Gateway | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Merge Gateway | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Vercel AI Gateway | google/gemini-3-flash | global | $0.50 | $3 | 1M | |
| OpenRouter | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Jiekou.AI | gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Perplexity Agent | google/gemini-3-flash-preview | global | $0.50 | $3 | 1M | |
| Venice AI | gemini-3-flash-preview | global | $0.70 | $3.75 | 256K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Available versions of this model. The score column identifies the version used in the overall ranking.
| Gemini 3 Flash | 64.8 | N/A | 1M | 65.5K | No | Proprietary |
Key information about Gemini 3 Flash and its available data.
Gemini 3 Flash offers frontier intelligence built for speed at a fraction of the cost. It combines Gemini 3's Pro-grade reasoning with Flash-level latency, efficiency and cost. Features a 1 million-token input context window and is optimized for agentic workflows, coding, and complex analysis.
Data as of 2026-08-11.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
Common questions about Gemini 3 Flash.
Gemini 3 Flash's default version was released on Dec 17, 2025.
Gemini 3 Flash's official API price is $0.50 per million input tokens and $3 per million output tokens via Google. The lowest tracked third-party offer starts at $0.07 input and $0.43 output via QiHang.
Gemini 3 Flash was created by Google.
The default version has a 1M token context window.
No. The default version is not marked as having publicly available weights.
22 provider offerings are linked to the default version.
Nearby ranked alternatives include Kimi K2.7 Code, Qwen3.5 397B A17B, Kimi K2.5.