Gemini API
Provider, cloud llm, by Google
Assessment
Wired as a third engine and rarely chosen: the local model covers most days and Claude the rest. Its long context and multimodal input are the reasons to pick it when they matter; otherwise it is a spare.
2026-10-09
Strengths
- Very long contexts and native multimodal input: documents, images, audio, video in one request
- A generous free tier for experiments
- Fast smaller models for high-volume, low-stakes calls
Limitations
- Rate limits and quota behaviour change often
- A third provider to keep wired when two already cover the need
Google's Gemini API serves the Gemini model family, notable for contexts of a million tokens and for taking text, images, audio and video in one request.[1] It has a free tier sized for development and a pay-as-you-go tier above it.
What it is for
Whole-document and whole-video questions, where the context length does the work. Experiments, where the free tier removes the cost question. A second cloud opinion when one is wanted.
Why it is rarely used
A setup with a capable home model for daily work and Claude for the hard cases has little left for a third engine to do, and each extra provider is another key, another quota and another set of quirks to keep working.[2] It stays wired for the day a task fits it.
Things to know
- Quotas and rate limits have changed several times; read the current table before relying on it in a scheduled job.
- The same egress rule applies: nothing goes to it without a reason.
History
- 2026-10-09: article written and published.
- 2026-10-09: entry created.
Comments
Public comments on each entry are coming. Nothing is collected here yet.