# Google launches Gemini 3.7 Flash and halves its price as Gemini 3.5 Pro waits > Source: > Published: 2026-08-14 02:12:20+00:00 *Google just shipped a second Gemini model in three weeks and cut its price in half, even as the flagship Gemini 3.5 Pro it promised in May still hasn't shown up.* Gemini 3.7 Flash went live on August 13, 2026, barely three weeks after Gemini 3.6 Flash. Three weeks. It's aimed at coding and autonomous agent work, and the benchmark jumps are hard to ignore: on DeepSWE v1.1, the model climbed from 49.0% to 65.3%, and on FrontierCode 1.1 Main it went from 34.4% to 43.6%. That's a real jump. Google says the model is better at debugging, producing deployable code on the first try, and handling the kind of multi-step tool calling that agentic workflows depend on. It's also cheaper, for now. Gemini 3.7 Flash costs $0.75 per million input tokens and $3.75 per million output tokens, half the standard rate, through the end of 2026. That won't last. Prices double to $1.50 and $7.50 on January 1, 2027, according to Google's own announcement covered by VentureBeat and TechTimes. The model rolled out first inside the Gemini app's Spark agent tier, and only for AI Pro and Ultra subscribers. Free users and AI Plus subscribers stay on Gemini 3.6 Flash for now, as 9to5Google reported. GitHub's changelog also lists it as newly available in Copilot for Pro, Pro+, Max, Business, and Enterprise users. ## A Flagship Gone Missing Here's the part that's hard to miss. Gemini 3.5 Pro, the model Google actually announced as its flagship at I/O back on May 19, still hasn't shipped. As of this week it's been delayed so many times that Tech Insider has taken to counting the days, putting the tally past 67 and calling it the longest-awaited model of 2026. Forbes reported the delay is still ongoing as of August 13, the same day 3.7 Flash landed. The holdup traces back to coding performance. According to 9to5Google, Google updated the training data behind Gemini in late June specifically to sharpen coding ability, and the results disappointed internally. Geeky Gadgets reported something more drastic: Google DeepMind scrapped the original Gemini 3.5 Pro build entirely after engineers found structural failures in recursive tool-calling and SVG generation. That's not a small bug. Recursive tool-calling is the exact skill an agentic coding assistant needs, chaining steps together without falling apart. A flagship model failing at it is a real setback, not a rounding error. So Google is doing what it can while the flagship gets rebuilt. Flash models are smaller and cheaper to train and ship. That makes them the natural place to patch a competitive gap fast. Frankly, shipping two Flash updates in three weeks looks less like a roadmap and more like triage. ## The Real Fight Is Price The competitive backdrop explains the urgency. OpenAI released GPT-5.6 Sol on July 9, which Sam Altman said was 54% more token efficient on agentic coding tasks. Grok 4.5 opened to the public the same day. Bloomberg framed Gemini 3.7 Flash's launch directly against that backdrop, noting it arrives while Google's actual top-tier model is still missing in action. Google's own benchmark tables, cited by VentureBeat, tell a flattering story: Claude Sonnet 5 at 10.7%, GPT-5.6 Terra at 23.6% on AutomationBench. Convenient numbers for Google. Anthropic and OpenAI aren't sitting still on price either. That's the real fight here. Developers building coding agents route based on cost per token as much as raw capability. Google cutting its Flash pricing in half through year-end is a direct bid to keep that traffic from drifting to Claude or GPT-5.6. Gemini 3.5 Pro, meanwhile, sits in the shop. None of this means Gemini 3.5 Pro is dead. It means Google decided it couldn't afford to wait for it. A three-week release cadence on a mid-tier model isn't how a company ships when things are calm. It's how a company ships when a competitor just took a bite out of its lunch and the main course isn't ready yet. **Also read:** [Nvidia Bets Its Balance Sheet That $500 Billion in AI Chips Won't Age](https://startupfortune.com/nvidia-bets-its-balance-sheet-that-500-billion-in-ai-chips-wont-age/) • [Meta's Smart Glasses Trigger a Privacy Backlash It Can't Fix With Software](https://startupfortune.com/metas-smart-glasses-trigger-a-privacy-backlash-it-cant-fix-with-software/) • [IBM Partners With OpenAI To Put GPT-5.6 Inside Its Consulting Business](https://startupfortune.com/ibm-partners-with-openai-to-put-gpt-56-inside-its-consulting-business/)