MOUNTAIN VIEW, California, August 16, 2026, 14:37 PDT
- Gemini 3.7 Flash costs $0.75 per million input tokens through year-end.
- Google cut both introductory input and output rates by 50%.
Google parent Alphabet (NASDAQ:GOOGL) has launched Gemini 3.7 Flash with introductory API prices half those of Gemini 3.6 Flash. The model targets coding and automated business workflows that require repeated tool calls.
The price cut matters most for long-running agents. Those systems plan tasks, query software and revise results across many steps. Token charges can accumulate quickly.
Google set the introductory rate at $0.75 per million input tokens. Output costs $3.75 per million tokens. Both offers run through the end of 2026.
| API price per million tokens | Gemini 3.7 Flash introductory rate | Gemini 3.6 Flash original rate | Reduction |
|---|---|---|---|
| Input | $0.75 | $1.50 | 50% |
| Output | $3.75 | $7.50 | 50% |
| Offer period | Through end-2026 | Original launch rate | Temporary for 3.7 |
A simple mixed workload now costs $4.50 for one million input and output tokens each. The equivalent 3.6 Flash bill was $9.00. Larger agent deployments preserve the same percentage saving.
| Illustrative workload | Gemini 3.7 Flash | Gemini 3.6 Flash original rate | Calculated saving |
|---|---|---|---|
| 1M input + 1M output tokens | $4.50 | $9.00 | $4.50 |
| 10M input + 2M output tokens | $15.00 | $30.00 | $15.00 |
| 100M input + 10M output tokens | $112.50 | $225.00 | $112.50 |
Gemini 3.7 Flash arrived only three weeks after version 3.6. Google says it improves debugging, issue resolution and production-ready code generation. Independent results across real workloads remain limited.
The model is available to developers through Google’s AI services. It also powers Gemini Spark, the company’s subscription agent for AI Pro and Ultra users in more than 160 countries.
| Google AI product | Status on August 16 | Primary role | Access detail |
|---|---|---|---|
| Gemini 3.7 Flash | Available | Coding and multi-step agent workflows | Developer services and Gemini Spark |
| Gemini 3.6 Flash | Available predecessor | General Flash workloads | Original API pricing was twice the 3.7 offer |
| Gemini 3.5 Pro | Still in partner testing | Premium flagship model | No launch date disclosed |
Google describes 3.7 Flash as its “most intelligent workhorse model yet.” The phrase signals a commercial priority: broad deployment at lower cost, rather than a new flagship benchmark leader. Tom’s Guide hands-on test
That distinction is important. Gemini 3.5 Pro remains in partner testing after earlier timing slipped. Google has not provided a release date for the premium model.
One early Spark test used Gmail and Drive to assemble deadlines, bills and appointments. It produced linked evidence and draft actions. It also missed files with unclear names, showing that workflow quality still depends on organization and prompts.
External rankings also temper the launch claims. An Artificial Analysis composite cited by MarketWatch placed Gemini 3.7 Flash ninth among current models. Such rankings vary by test mix and can change quickly.
The pricing therefore carries the clearest measurable change. A 50% cut can make retries, tool calls and verification steps cheaper. It does not prove that agents will finish tasks correctly.
Businesses should compare total workflow cost, not token rates alone. Search, storage, databases, external APIs and human review can exceed the model bill.
Risks: The discounted prices are introductory and may change after 2026. Google has not published enough independent production evidence to verify reliability, security or savings across complex enterprise agents.
Related articles: News
