Google DeepMind

It builds the Gemini frontier model family plus the open-weight Gemma models, Veo (video), Lyria (music), Imagen, and the Gemini Robotics ER embodied-reasoning line. Models ship through two commercial surfaces: the Gemini Developer API / Google AI Studio (ai.google.dev, self-serve API keys) and the enterprise Vertex AI / Gemini Enterprise Agent Platform on Google Cloud; the positioning is long-context (1M-token), natively multimodal, tool-integrated models at aggressive price-performance versus OpenAI and Anthropic.

A Access and limits

API keys are self-serve from Google AI Studio (aistudio.google.com) with no credit card required for the free tier. FREE TIER: most Gemini text models (3.8/3.7/3.6/3.5 Flash, 3.5 & 3.1 Flash-Lite, 3 Flash Preview, 2.5 Pro/Flash/Flash-Lite, TTS, Live, Transcribe, embeddings, robotics, Gemma 4) are free of charge at limited rates; free-tier content IS used to improve Google's products, while paid-tier content is not. Notably NOT on the free tier: Gemini 3.1 Pro Preview, all Nano Banana image models, Gemini Omni Flash, 2.5 Pro TTS, 2.5 Computer Use, Veo and Lyria.

Google no longer publishes free-tier RPM/TPM/RPD in the docs — per-model live limits are shown at aistudio.google.com/rate-limit. PAID USAGE TIERS: Free (active project) -> Tier 1 (link an active billing account, $250 billing cap) -> Tier 2 (paid $100 + 3 days, $2,000 cap) -> Tier 3 (paid $1,000 + 30 days, $20,000-$100,000+ cap); a rolling 10-minute spend cap also applies ($10 / $50 / $200 for Tiers 1/2/3). CONSUMPTION MODES: Standard; Batch API and Flex inference at 50% of standard; Priority inference at 180% of standard (1.8x) with 0.3x the standard rate limit.

CONTEXT-LENGTH TIERS: Gemini 3.1 Pro Preview, 2.5 Pro and 2.5 Computer Use are cheaper for prompts <=200k tokens and step up above 200k — the recorded input/output prices are the <=200k base tier. INTRODUCTORY PRICING: Gemini 3.8/3.7/3.6 Flash are at half price ($0.75/$3.75) only through December 31, 2026; standard $1.50/$7.50 begins January 1, 2027. TOOLS ARE BILLED SEPARATELY: Google Search grounding on Gemini 3.x models is 5,000 free requests/month shared across all Gemini 3.x models, then $14 per 1,000 requests (Gemini 2.5 models: 1,500 RPD free, then $35 per 1,000 grounded prompts); Google Maps grounding $25 per 1,000 grounded prompts after free quota; File Search embeddings $0.15/1M tokens; code execution and URL context are billed as ordinary model tokens.

Context-cache storage is billed by the hour on top of cached-read tokens ($1.00/1M tokens/hour for most Flash models, $4.50/1M/hour for Pro-class and 3.8/3.7/3.6 Flash after the intro period, $0.50/1M/hour during the intro period). VERTEX AI: the same models are sold on Vertex AI / Gemini Enterprise Agent Platform at identical global-endpoint prices, but non-global (regional) endpoints carry roughly a 10% premium (e.g. Gemini 3.8 Flash $0.825 input non-global vs $0.75 global); Vertex adds provisioned throughput, FSP committed-use discounts, VPC-SC and data-residency controls.

Batch API is capped by enqueued tokens per model per tier (e.g. Tier 1: 3M tokens for 3.8 Flash, 10M for 3.5 Flash-Lite; Tier 2: 400M-500M), with 100 concurrent batch jobs, 2GB input file limit and 20GB file storage.

C Calling the API

Google DeepMind accepts requests in the OpenAI Chat Completions format, so any OpenAI-compatible client works by changing the base URL. Example uses gemini-embedding-001.

curl https://generativelanguage.googleapis.com/v1beta/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-embedding-001",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

from openai import OpenAI

client = OpenAI(
    base_url="https://generativelanguage.googleapis.com/v1beta",
    api_key=os.environ["API_KEY"],
)

response = client.chat.completions.create(
    model="gemini-embedding-001",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

A compatibility layer is not always the provider's full API — caching, strict tool schemas and structured output are commonly unsupported there. Check the official docs before relying on an advanced feature.

D Official references

• Homepage
• API documentation
• Pricing