Gemini 3.7 Flash Launch: Pricing, Specs & Evals (2026)
Google and Google DeepMind officially launched Gemini 3.7 Flash on August 13, 2026. Released just three weeks after Gemini 3.6 Flash, Google describes Gemini 3.7 Flash as its “most intelligent workhorse model yet for coding, agentic reasoning, and complex developer workflows.”
Logan Kilpatrick and Sundar Pichai confirmed the release across developer channels, highlighting core algorithmic improvements to the model’s reasoning foundation, customizable thinking configurations, and a dramatic 50% introductory price reduction for developers and enterprise teams.
Here is the complete, official breakdown of Gemini 3.7 Flash: confirmed pricing, full benchmark comparison tables from Google DeepMind, availability surfaces, and early developer reactions.
1. Confirmed Pricing and Model Specifications
Google is offering promotional introductory pricing for Gemini 3.7 Flash through December 31, 2026:
| Metric | Introductory Rate (Through Dec 31, 2026) | Standard Rate (Starting Jan 1, 2027) | Notes |
|---|---|---|---|
| Input Tokens | $0.75 per 1M tokens | $1.50 per 1M tokens | 50% price reduction vs 3.6 Flash |
| Output Tokens | $3.75 per 1M tokens | $7.50 per 1M tokens | 50% price reduction vs 3.6 Flash |
| Context Window | 1,000,000 tokens (1M) | 1,000,000 tokens (1M) | Standard 1M token input buffer |
| Max Output | 65,536 tokens | 65,536 tokens | Supports large single-pass code generation |
| Model ID | gemini-3.7-flash |
gemini-3.7-flash |
Production API identifier |
Availability Across Surfaces
- Developers: Available immediately in Gemini API / Google AI Studio, Vertex AI, Antigravity IDE, and Android Studio.
- Enterprises: Deployed to Gemini Enterprise Agent Platform and Gemini Enterprise app.
- Consumers: Powers Gemini Spark for Google AI Pro and Ultra subscribers across 160+ supported countries.
2. Official Benchmark Performance (Google DeepMind)
The official model card published by Google DeepMind details significant leaps in software engineering, multi-step agent automation, and legal document comprehension compared to Gemini 3.6 Flash and competing frontier models:
| Benchmark Category | Gemini 3.7 Flash | Gemini 3.6 Flash | Claude Sonnet 5 | GPT-5.6 Terra | Muse Spark 1.2 |
|---|---|---|---|---|---|
| Artificial Analysis Intelligence Index | 56 | 52 | 55 | 57 | 57 |
| FrontierCode 1.1 Main (Production Code) | 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 / WebDev Arena (Elo Rating) | 1588 | 1538 | 1541 | 1523 | 1535 |
| Terminal-Bench 2.1 (Agentic CLI Tasks) | 85.8% | 78.0% | 80.4% | 87.4% | 82.9% |
| AutomationBench (Enterprise Workflow Automation) | 30.4% | 17.0% | 10.7% | 23.6% | - |
| GDPVal-AA v2 (Knowledge Work Elo) | 1525 | 1422 | 1598 | 1578 | 1628 |
| Harvey LAB-AA (Legal Document Analysis) | 90.7% | 85.1% | 90.1% | 85.2% | - |
| GDP.pdf (Complex PDF Comprehension) | 34.0% | 22.0% | 28.0% | 24.7% | 16.0% |
| LVBench (Long Video Comprehension) | 85.4% | 84.2% | 68.5% | 78.9% | - |
| GDM-MRCR v2 (8-Needle 128k Retrieval) | 97.0% | 91.8% | 81.5% | 93.5% | - |
| OSWorld-2.0 (Agentic Computer Use) | 47.9% | 33.8% | - | 50.2% | - |
Source: Official Google DeepMind Gemini 3.7 Flash Model Card (August 13, 2026).
Key Benchmark Takeaways
- DeepSWE v1.1 Jump: Gemini 3.7 Flash improved by +16.7 percentage points over 3.6 Flash (65.3% vs 48.6%), surpassing Claude Sonnet 5 on complex long-horizon repository editing.
- AutomationBench: Massive jump to 30.4% (up from 17.0%), outperforming GPT-5.6 Terra (23.6%) and Claude Sonnet 5 (10.7%) on multi-step enterprise workflows.
- Terminal-Bench 2.1: Achieved 85.8% task completion, placing it within 1.6 points of GPT-5.6 Terra while operating at a fraction of the token cost.
3. Key Feature Upgrades and Capability Improvements
1. Customizable Thinking Configurations
Developers can adjust Gemini 3.7 Flash’s reasoning depth dynamically depending on the task:
- Low Thinking / Fast Mode: Optimized for real-time autocomplete, UI layout generation, and rapid chat responses.
- Deep Thinking Mode: Allocates additional internal reasoning tokens for complex debugging, mathematical proofs, and multi-file code refactoring.
2. Multi-Step Agentic Discipline
Google emphasized improved instruction following and roadblock adaptation. The model shows stronger discipline in tool calls, reducing hallucinated parameters during API execution and RAG lookups.
3. Native Multimodal Input Capabilities
Supports text, images, video, audio, and PDF parsing out of the box. Output support includes structured JSON output, native code execution, system instructions, and preview access for Computer Use agents.
4. Early Community Reactions and Developer Sentiment
Following the 17:04 UTC launch, initial reaction across developer forums, X, and OpenRouter has focused heavily on cost-efficiency at scale:
- Price-to-Performance Leadership: Developers operating autonomous agent swarms highlighted that at $0.75 / $3.75 per 1M tokens, Gemini 3.7 Flash is 60% to 70% cheaper than GPT-5.6 Terra and Claude Sonnet 5 while scoring within 1-2 points on major coding benchmarks.
- Rapid 3-Week Cadence: The engineering community praised the tight 3-week iteration window from 3.6 to 3.7 Flash, signaling Google’s aggressive focus on high-volume developer infrastructure.
- Internal Google Usage: Sundar Pichai noted positive internal adoption by Google engineers using Gemini 3.7 Flash within Antigravity for production code generation.
5. Frequently Asked Questions
Is Gemini 3.7 Flash officially launched?
Yes. Google officially launched Gemini 3.7 Flash on August 13, 2026. It is available immediately in Google AI Studio, Vertex AI, Gemini API, and supported consumer surfaces.
What is the price of Gemini 3.7 Flash?
Introductory pricing through December 31, 2026 is $0.75 per 1M input tokens and $3.75 per 1M output tokens (a 50% discount compared to Gemini 3.6 Flash). Standard pricing ($1.50 / $7.50) resumes on January 1, 2027.
What is the model ID for API integration?
The official API model identifier is gemini-3.7-flash.
How does Gemini 3.7 Flash compare to GPT-5.6 Terra?
On the Artificial Analysis Intelligence Index, Gemini 3.7 Flash scores 56 vs GPT-5.6 Terra’s 57. However, Gemini 3.7 Flash is significantly cheaper ($0.75/$3.75 introductory vs $5/$30), while outperforming GPT-5.6 Terra on GDP.pdf document parsing (34.0% vs 24.7%) and FrontierCode 1.1 (43.6% vs 41.3%).
Last updated: August 13, 2026 (17:15 UTC). Data sourced directly from official Google DeepMind model cards and Google AI announcements.
Official Primary Sources:
- Google Official Launch Blog: blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash
- Google DeepMind Gemini 3.7 Flash Model Card: deepmind.google/models/model-cards/gemini-3-7-flash
- Google DeepMind Evals Methodology: deepmind.google/models/evals-methodology/gemini-3-7-flash
Related articles
AI Assistant Hacks Gym Website: What Happened in Australia
What reporting says about an AI assistant using a gym booking API to move a user up a waitlist—and why the incident is a warning about agent permissions.
ChatGPT GPT-5.6 Sol Update: Access, Reasoning, and Factuality Changes (August 2026)
Side-by-side breakdown of Free/Go vs Plus/Pro tiers after OpenAI’s 6 August 2026 ChatGPT update. Pricing, unlimited Luna access, reasoning slider, and factual error reduction.