Google released Gemini 3.7 Flash today at $0.75 per million input tokens and $3.75 per million output tokens, half the launch price of Gemini 3.6 Flash and just three weeks after that model shipped. The company is pitching 3.7 Flash as its most capable workhorse yet for coding and agent workloads, with double-digit gains on nearly every benchmark Google chose to publish. The compressed release cadence, and the price cut that came with it, are the story.
On DeepSWE v1.1, 3.7 Flash scores 65.3% against 49.0% for 3.6 Flash — a 16-point jump on a software-engineering eval in under a month. FrontierCode 1.1 Main rises to 43.6% from 34.4%. On Arena.ai's WebDev Arena, the model posts a 1588 Elo, up from 1538. Google says the gains came from developer feedback and algorithmic changes rather than a larger model.
“Today, we're building on the progress of our widely used Flash series by introducing Gemini 3.7 Flash, our most intelligent workhorse model yet for coding and agents.”— Tulsee Doshi, Senior Director, Product Management, Google
The reasoning benchmarks moved even more sharply. AutomationBench, which measures real-world business workflow completion, nearly doubles to 30.4% from 17.0%. GDP.pdf, an eval for reading complex documents, climbs to 34.0% from 22.0%. Google is positioning these numbers at knowledge-dense fields — finance, law, biosciences — where Flash-tier pricing has historically been a hard sell against the larger Pro and Ultra tiers.
Key facts
- 01Gemini 3.7 Flash launches three weeks after 3.6 Flash at $0.75/1M input tokens and $3.75/1M output tokens — half the prior price.
- 02The model scores 65.3% on DeepSWE v1.1, up from 49.0% for 3.6 Flash, and 43.6% on FrontierCode 1.1 Main versus 34.4%.
- 03On Arena.ai's WebDev Arena, 3.7 Flash posts a 1588 Elo, up from 1538 for the prior generation.
- 04AutomationBench score nearly doubles to 30.4% from 17.0%, and GDP.pdf reasoning jumps to 34.0% from 22.0%.
- 05Gemini Spark, the 24/7 personal agent for Google AI Pro and Ultra subscribers in 160 countries, switches to 3.7 Flash today.
The pricing move is the sharper signal. Halving input and output costs against a three-week-old model, and calling it an introductory rate that holds through the end of the year, tells developers Google is willing to trade near-term margin for agent-workload volume. Autonomous agents burn tokens by the millions on multi-step tool calls, and a 2x cost reduction directly changes which workflows pencil out.
Tulsee Doshi, senior director of product management on the Gemini team, framed 3.7 Flash as a direct response to developer feedback, particularly on agent reliability. Google says the model adapts better to roadblocks, asks for clarification when intent is ambiguous, and follows instructions with higher fidelity across long tool-calling chains.
“A more disciplined execution means less manual oversight and fewer retries across engineering workflows.”— Tulsee Doshi, Senior Director, Product Management, Google
The improvements flow into Gemini Spark, the 24/7 personal agent Google launched at I/O for Google AI Pro and Ultra subscribers. Spark is now available in more than 160 countries and switches to 3.7 Flash starting today. Google says the upgrade sharpens Spark's tool use inside Google Workspace — consolidating files, drafting emails, and updating status documents — where prior Flash models struggled with multi-skill sequencing.
For developers, 3.7 Flash lands in the Gemini API through Google AI Studio and Android Studio, and in Google Antigravity for agent-first workflows. Enterprises get access through the Gemini Enterprise Agent Platform and the Gemini Enterprise app. The company is also shipping updated Frontier Safety safeguards covering chemical, biological, radiological, nuclear, and cyber-offense misuse categories.
The three-week gap between 3.6 and 3.7 Flash is the aggressive part of this launch. Rapid iteration at that cadence risks confusing enterprise procurement teams, who are still evaluating whether to build on last month's Flash release. It also raises the question of how quickly a 3.8 Flash arrives, and whether the introductory price survives contact with a follow-on model that renders these benchmarks stale. Google has not addressed either question.
For the broader model market, 3.7 Flash sets a new floor on cost-per-capability at the workhorse tier. A model that scores 65.3% on DeepSWE at $0.75 per million input tokens compresses the pricing envelope that OpenAI, Anthropic, and every open-weights provider are working against. The next round of agent-platform pricing announcements — from anyone shipping something that runs 24/7 on someone else's tokens — will have to answer this number.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
