model release

Google Releases Gemini 3.7 Flash With 1M-Token Context and Multimodal Input

TL;DR

Google has released Gemini 3.7 Flash, a multimodal model built for agentic workflows, coding, and multi-step reasoning. It offers a 1,049K token context window and is priced at $0.38 per million input tokens and $1.88 per million output tokens, available now via OpenRouter.

2 min read
0

Gemini 3.7 Flash at a glance

Context window1049K tokens
Input$0.75/1M tokens
Output$3.75/1M tokens

How Gemini 3.7 Flash ranks now

BenchmarkScoreRankSource
Arena Elo1490 Elo15 of 146LMArena, Nov 2026 ↗
GPQA94.5%7 of 165Artificial Analysis, Aug 2026 ↗

Rank is against every model in our database with a sourced score on that benchmark, each counted once with its best score.

All Gemini 3.7 Flash specs and scores →

Google has released Gemini 3.7 Flash, a multimodal AI model positioned for fast agentic workflows, coding tasks, and complex multi-step reasoning. The model is now accessible through OpenRouter's API under the identifier google/gemini-3.7-flash.

Specifications

Gemini 3.7 Flash ships with a context window of 1,049,000 tokens, placing it among the largest-context models currently available. According to Google, the model is designed for tasks requiring responsive performance and reliable execution across multiple sequential steps — a profile aimed squarely at agentic and coding use cases rather than single-shot chat.

The model accepts text, image, file, audio, and video inputs and produces text outputs, making it a fully multimodal system rather than a text-only release.

Pricing

OpenRouter lists pricing for Gemini 3.7 Flash at $0.38 per million input tokens and $1.88 per million output tokens. That output price is roughly five times the input price, a common pattern for models tuned to reward concise, high-value generations over lengthy ones.

What's known and what isn't

Google's own description frames Gemini 3.7 Flash around agentic workflows, coding, and multi-step reasoning, but no independent benchmark scores were included in the release materials reviewed for this article. No parameter count, training data cutoff date, or detailed architecture information has been disclosed. Claims about the model's suitability for agentic tasks and coding come directly from Google's positioning and have not been independently verified with benchmark data at time of writing.

What this means

A 1,049K-token context window paired with sub-$2 output pricing signals Google is continuing to push the Flash line as a high-volume, cost-efficient option for developers building agents and coding tools that need to process large amounts of context — think entire codebases, long documents, or extended video/audio inputs — without paying frontier-model prices. The five-to-one output-to-input pricing ratio suggests Google expects most usage to be input-heavy (large context, short responses), which aligns with retrieval-augmented and agentic patterns where a model reads a lot but writes comparatively little.

The absence of published benchmark numbers makes it difficult to assess how Gemini 3.7 Flash compares to competing fast-tier models on coding or reasoning tasks. Developers evaluating it for production agentic pipelines will need to run their own tests until third-party benchmarks emerge. Availability via OpenRouter at launch also means the model is immediately accessible to a broad developer base without requiring direct Google Cloud onboarding.

Comments

Loading...