Gemini 3 API — pricing, availability, and the live way to call Gemini today

Gemini 3 is Google's current model family — Gemini 3 Pro and Gemini 3.1 Pro at the top, and a Flash line that runs from Gemini 3 Flash to Gemini 3.8 Flash — natively multimodal, with 1M-token context and chain-of-thought reasoning. You can call the Gemini 3 API on Kunavo today through one OpenAI-compatible endpoint: Gemini 3.1 Pro at $0.70 / $4.20 per 1M tokens, about 65% under Google's $2.00 / $12.00 list, and Gemini 3.8 Flash, the newest Flash, at $0.525 / $2.625 — 30% under the introductory $0.75 / $3.75 Google charges for 3.x Flash through December 31, 2026. This guide covers what every tier costs, which slugs are live, and the code to call them.

Gemini 3 API pricing

Google's official rates for the family, next to what each tier costs on Kunavo. For the 3.x Flash models Google charges an introductory rate until December 31, 2026, so the table shows that rate — the one you would pay by going direct today — alongside the standard rate that follows it:

Every current Gemini model's Kunavo rate next to Google's, plus worked cost examples, is in the Gemini API pricing guide.

Gemini 3.8 Flash API: price and Google's intro rate

Gemini 3.8 Flash is Google's newest Flash, released September 2, 2026 and aimed at long-horizon software engineering and agentic execution, with a 1,048,576-token context window and up to 65,536 output tokens. On Kunavo the Gemini 3.8 Flash API costs $0.525 input / $2.625 output per 1M tokens, called as gemini-3-8-flash on the same endpoint as every other model.

Google charges an introductory $0.75 / $3.75 per 1M tokens for its 3.x Flash models — 3.6, 3.7 and 3.8 Flash — through December 31, 2026, and its standard $1.50 / $7.50 from January 1, 2027. Against the rate Google charges today, Kunavo's Gemini 3.8 Flash and Gemini 3.7 Flash ($0.525 / $2.625) are 30% cheaper. Gemini 3.6 Flash is not: at $1.05 / $5.25 it costs 40% more than Google's intro rate, so 3.7 and 3.8 Flash are the ones to use — both newer and cheaper.

What Kunavo adds is one OpenAI-compatible key and one prepaid balance across Gemini, Claude, GPT and the image and video models, with no Google Cloud project and no charge for failed requests. What it does not add is dedicated capacity: Kunavo is shared capacity with no dedicated quota and no contractual SLA, so if you need either, call Google directly.

Gemini 3.6 Flash — what changed

Gemini 3.6 Flash (released July 21, 2026) was the first 3.x Flash we could serve, and the model behind the reasoning-token measurements in this section. Google's standard rate for it is $1.50 input / $7.50 output per 1M tokens — at standard rates the same input price as Gemini 3.5 Flash with output cut about 17% — but through December 31, 2026 Google charges an introductory $0.75 / $3.75. Google pitches it at long-horizon agent work, where it claims to cut token cost substantially on multi-step engineering tasks.

The number that actually decides your bill is not on the price list: 3.6 Flash thinks by default, and the thinking dominates the output meter. In our testing a one-sentence factual answer spent roughly 120–140 reasoning tokens to produce 7 visible output tokens. Reasoning tokens bill at the output rate, so the honest cost of a short 3.6 Flash answer is closer to 150 output tokens than to 7 — roughly 20× the naive estimate. Two practical consequences:

• Do not set a small max_tokens. Anything under ~150 gets consumed by reasoning before the model emits an answer, and you receive a truncated fragment while still paying for the thinking.
• Budget on reasoning, not on visible output. For short, high-volume calls a non-thinking Flash tier is usually cheaper in practice, even at a higher sticker price per token.

Gemini 3.6 Flash is still live on Kunavo as gemini-3-6-flash at $1.05 / $5.25 per 1M, but that is 40% above the $0.75 / $3.75 Google charges until December 31, 2026, and above Kunavo's own price for the newer models. For new work call gemini-3-8-flash or gemini-3-7-flash at $0.525 / $2.625, 30% under Google's intro rate. They are thinking models too, so keep max_tokens generous on them as well.

Call Gemini on Kunavo today

Kunavo fronts Gemini behind the OpenAI-compatible /v1/chat/completions endpoint — keep the OpenAI SDK, change base_url, no Google Cloud project. This calls the Gemini 3.8 Flash API; swap the model string for gemini-3-7-flash, gemini-3-1-pro or gemini-2-5-flash and nothing else changes:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["KUNAVO_API_KEY"],
    base_url="https://api.kunavo.com/v1",
)

# Gemini 3.8 Flash is Google's newest Flash. Mind the max_tokens: the 3.x Flash
# models think before they answer and reasoning bills as output, so a small
# cap can be spent thinking and return a truncated answer.
resp = client.chat.completions.create(
    model="gemini-3-8-flash",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Summarize this contract in 5 bullets."}],
)
print(resp.choices[0].message.content)

Getting a key takes a minute — see the Gemini API key guide. The same sk-kn- key also reaches Claude, GPT, Nano Banana image and Veo 3 video.

Which Gemini tier should you call?

• Long-context and reasoning workloads — gemini-3-1-pro at $0.70 in / $4.20 out. It is Google's flagship tier for coding, agents and cross-modal analysis, and the Pro tier we serve.
• Agentic work that reasons per step — gemini-3-8-flash at $0.525 / $2.625, 30% under Google's intro rate, or gemini-3-7-flash at the same price if your prompts are validated there. Reasoning at Flash latency; budget on reasoning tokens, not visible output.
• High-volume, latency-sensitive calls — gemini-2-5-flash at $0.09 in / $0.75 out remains the cost-efficiency play; it does not reason by default, which is exactly why it stays cheap on short answers.
• Not 3.6 Flash, on price — gemini-3-6-flash at $1.05 / $5.25 costs more than 3.7 and 3.8 Flash on Kunavo and more than Google's own intro rate; keep it only for prompts already validated on it.
• Building for the next upgrade — code against the OpenAI-compatible shape and treat the model string as config; new Gemini slugs land behind the same endpoint.

FAQ