Google lists Gemini 3.7 Flash with 1M-token context, 65,536-token output and preview computer use
- Gemini 3.7 Flash accepts text, images, video, audio, and PDFs, returns text, and has a 1,048,576-token input limit with a 65,536-token output limit.
- Google lists support for code execution, function calling, file search, Search and Google Maps grounding, structured outputs, URL context, and caching.
- The model supports computer use in preview, while audio generation, image generation, and the Live API are not supported.
- Thinking supports low, medium, and high levels; requesting the minimal level returns an error.
- Google lists
gemini-3.7-flashas the stable model version and says it supports Batch API, Flex inference, and Priority inference.
Hacker News opinions
December 2026 introductory pricing makes me think there will be no major Flash development before next year.
I read the price as a competitive move, not a roadmap signal. Flash 3.6 shipped only three weeks ago, so Google could still ship 3.8 before then.
Gemini 4 is reportedly close, so a new Flash update should arrive unless that release slips by months.
Cognition's FrontierCode puts Terra at roughly half the price, and Grok 4.6 looks both better and cheaper. I do not see why I would choose 3.7 Flash under those conditions.
I still use 3.6 Flash for token speed, latency, uptime, and long-context work. Grok 4.6 scored lower on some of my internal tests or ran slower.
Artificial Analysis estimates Grok 4.6 costs 485 for Gemini 3.7 Flash. Per-token prices miss how many tokens each model spends to finish a task.
For my workload, Flash 3.6 High was about 10 times faster than Luna xhigh and got similar results.
I want another Gemini Pro model. Google seems focused on fast, moderately capable models instead.
The model card places it around GPT 5.6 Tera and Sonnet 5, depending on the benchmark.
I think this is competitive because it beats Claude Sonnet 5 on most benchmarks at less than half the price.
Sonnet 5 is a poor benchmark for cost effectiveness. I care more about how this compares with Luna, although 3.7 Flash at least looks usable and very fast.
For text-only work, DS V4 Flash or Pro is 13 to 26 times cheaper with comparable intelligence across many inference providers. Gemini's main advantage looks like speed, likely from TPUs, plus multimodal input.
My testing says Gemini 3.5 and 3.6 Flash are better than DS V4 Flash and Pro at text tasks.
Artificial Analysis has 3.7 Flash at 56 versus 52 for 3.6, but output per task rose from 26k to 37k tokens. Its halved price still makes 3.7 cheaper per task.
Google says prices double on January 1, 2027, to 7.50 per million output tokens. I doubt anyone will want a 3.7 model then, given how quickly newer models arrive and how many users moved away after earlier Gemini price hikes.
I spent at least 0 on 3.5, 3.6, and 3.7 after Google's price hike. Lower prices now do not matter much to me because alternatives are plentiful.
The delayed increase may be designed to move users to newer models as soon as they are available.
I would rather know a price hike in advance if I were building a product around the model.