Gemini 3.7 Flash
Best for tackling complex agentic tasks at scale
Our most intelligent workhorse model yet for coding and agents.
Intelligence in a Flash
Advanced reasoning at Flash-level latency and scale.
Slide 1 of 4
Versatility across agentic tasks
Optimized for high performance across software engineering, web development and knowledge work tasks.
Smart and rigorous
Rigorous reasoning efforts for better quality output.
Reliable in agentic execution
Navigates roadblocks and resolves coding and real world issues with accuracy.
Truly multimodal
Multimodal understanding across text, audio, images, code, and video.
Hands-on
Here are a few ways you can use 3.7 Flash’s improved agentic coding and knowledge work capabilities.
Slide 1 of 4
3D game built in Google Antigravity
From a simple text prompt to a fully playable 3D game. We used Gemini 3.7 Flash combined with Nano Banana to dynamically generate characters, items, and textures in real-time.
Interactive parallax landing pages
Stunning, interactive landing pages generated in a single shot. We used Gemini 3.7 Flash to orchestrate sub-agents, using Gemini Omni to create smooth, interactive parallax components.
Train a robotics model with multimodal graph looping
Watch a robotics model learn faster. We combined Gemini 3.7 Flash's multimodal understanding with a 3-agent graph loop to speed up the training loop.
Interactive annual report
From a static PDF to an interactive data story. Watch how complex annual reports are transformed into engaging web experiences complete with live charts and aggregated insights.
Showcase
Slide 1 of 12
“On Box's evaluation of real enterprise knowledge work, Gemini 3.7 Flash was both more accurate and significantly faster than the prior model, with its largest gains on the most challenging analytical tasks. This is the kind of progress that expands what AI can take on for organizations.”
Yashodha Bhavnani, VP of AI Products, Box
“The Gemini 3.7 Flash agent was 35% cheaper than 3.6 Flash, with a +8% observed prompt-cache hit rate and fewer tool errors.”
Gregor Zunic, Co-founder & CTO, Browser Use
“We’re excited by how well the new Gemini Flash model understands motion across time. For Cartwheel, that means more than recognizing what’s in a frame—it means following an action and reasoning about whether the motion is actually useful. We saw promising results on both fronts.”
Andrew Carr, CoFounder and Chief Scientist, Cartwheel
“Until now, getting high-quality answers from complex enterprise data has been expensive at scale. Models like Gemini 3.7 Flash are changing that by delivering better intelligence at dramatically lower cost. With Gemini 3.7 Flash on Databricks AI Gateway, enterprises can ask big questions of their data, like predicting next week’s revenue, identifying key customers, or deciding where to invest next.”
Ivan Zhou, AI Research Manager, Databricks
“Gemini 3.7 Flash’s multimodal capabilities shine on Emergent, accurately translating design inspiration and brand guidelines into full-stack web and mobile applications. It has never been easier for anyone, from small business owners to entrepreneurs, to bring their ideas to life.”
Madhav Jha, Co-Founder and CTO, Emergent
“Gemini 3.7 Flash is a significant improvement over prior Flash models on legal work. Based on our early testing, the model lifts all-pass by 2.6 pts compared to Gemini 3.6 Flash on Legal Agent Bench, with broad gains in quality across practice areas.”
Niko Grupen, Head of Applied Research, Harvey
"The new Gemini Flash model delivered Opus-level grounding performance, nearly matching information recall with materially higher precision. Combined with faster and more token-efficient outcomes, it makes for a very intriguing model."
Adithya Ramanathan, Head of Agent Development & Research, Hebbia
“On a subset of Harbor-Index run inside our DeepAgents harness, the model scored in the same band as the strongest models we tested, and it was the only model in our comparison to make progress on some tasks that defeated all the others. For a model in this tier, that is an excellent result on multi-step agentic work.”
Nick Hollon, Software Engineer, LangChain
“Gemini 3.7 Flash achieves on-par performance with mid-sized models such as GPT-5.6-Terra on our benchmarks, but at around half the cost, making it a strong contender in that category.”
Kyrill Hux, Co-Founder, Nunu.ai
“Gemini 3.7 Flash very efficiently implemented and audited its work on tasks porting Figma designs into code, with precise visual parity that other frontier models struggle with. I'm genuinely in a design 'psychosis' moment with this model. It just works.”
Luke Parker, Senior Software Engineer, OpenCode
“Google keeps making the strongest case that multimodal agents are ready for production — with Gemini 3.7 Flash, sharing video with a model becomes both natural and dramatically more efficient. It's supported in Pydantic AI from day one, alongside Gemini Live, which powers our new realtime agents.”
Douwe Maan, Lead Developer, Pydantic AI
“On a complex neurodegenerative disorder study experimental plan, I found Gemini 3.7 Flash's answer original and remarkably consistent with the latest experimental data. Compared to other models, this is the most compelling and accurate answer I have received coming very very close to what I see in the lab."
Piere Rodriguez Aliaga Ph.D., Basic Life Research Scientist, Stanford Department of Biology
Performance
3.7 Flash delivers better intelligence for complex workflows
| Benchmark | Notes | Gemini 3.7 Flash | Gemini 3.6 Flash | Claude Sonnet 5 | GPT-5.6 Terra | Muse Spark 1.2 |
|---|---|---|---|---|---|---|
| Input price $/1M tokens | $0.75* | $0.75* | $2.00 | $2.00 | $1.25 | |
| Output price $/1M tokens | $3.75* | $3.75* | $10.00 | $12.00 | $4.25 | |
| Artificial Analysis Intelligence Index Composite model intelligence | 56 | 52 | 55 | 57 | 57 | |
| FrontierCode 1.1 Main Production code quality | Score | 43.6% | 34.4% | 42.7% | 41.3% | — |
| DeepSWE v1.1 Long-horizon software engineering | 65.3% | 48.6% | 53.8% | 69.6% | 54.9% | |
| Code Arena Web development | Elo | 1588 | 1538 | 1541 | 1523 | 1535 |
| Terminal-bench 2.1 Agentic terminal coding | 85.8% | 78.0% | 80.4% | 87.4% | 82.9% | |
| Terminal-bench 3.0 General agent capabilities | 14.9% | 5.4% | 14.6% | 20.8% | — | |
| AutomationBench Enterprise workflow automation | Private set | 30.4% | 17.0% | 10.7% | 23.6% | — |
| GDPVal-AA v2 Knowledge work | Elo | 1525 | 1422 | 1598 | 1578 | 1628 |
| Harvey LAB-AA Complex legal workflows | 90.7% | 85.1% | 90.1% | 85.2% | — | |
| GDP.pdf Expert PDF document comprehension | 34.0% | 22.0% | 28.0% | 24.7% | 16.0% | |
| CharXiv Reasoning Information synthesis from complex charts | No tools | 84.5% | 85.2% | 77.0% | 85.9% | — |
| With tools | 88.7% | 89.4% | 88.3% | — | — | |
| LVBench Long video understanding | 85.4% | 84.2% | 68.5% | 78.9% | — | |
| GDM-MRCR v2 (8-needle) Long context performance | 128k (average) | 97.0% | 91.8% | 81.5% | 93.5% | — |
| OSWorld-2.0 Agentic computer use | 47.9% | 33.8% | — | 50.2% | — | |
| Agent's Last Exam Multimodal desktop and OS agent tasks | Pass rate | 26.3% | 24.2% | 33.3% | 28.0% | — |
| HLE-Verified Multidisciplinary expert reasoning | 53.6% | 51.2% | 31.0% | 51.1% | — | |
| BioMysteryBench Bioinformatics research reasoning | Human solvable | 87.1% | 80.6% | 87.5% | 83.8% | — |
| Human difficult | 43.5% | 41.2% | 34.1% | 49.4% | — | |
| LABBench2 Biology real-world research tasks | 82.1% | 76.1% | 80.1% | 81.2% | — |
Methodology: deepmind.com/models/evals-methodology/gemini-3-7-flash
* For 3.6 and 3.7 Flash, introductory price expires on December 31, 2026. Starting January 1, 2027, $1.50/1M input tokens and $7.50/1M output tokens will apply.
Model information
- Name
- 3.7 Flash
- Status
- General availability
- Input
- Text
- Image
- Video
- Audio
- Output
- Text
- Input tokens
- 1M
- Output tokens
- 64k
- Tool use
- Function calling
- Search as a tool
- Computer use
- Best for
- Everyday tasks
- Agentic coding
- Advanced reasoning
- Multimodal understanding
- Knowledge work
- Availability
- Gemini App
- Gemini Enterprise App
- Gemini Enterprise Agent Platform
- Google AI Studio
- Gemini API
- Google Antigravity
- Documentation
- View developer docs
- Model card
- View model card