Gemini API Pricing 2026: 4 Argon, 3.8 Flash, 3.1 Pro and Flash-Lite
Gemini 4 Argon's price doubles on 1 December and 3.8 Flash's on 1 January. Here's what every Gemini text model costs direct from Google, through Requesty and OpenRouter, and on SayGM, where 3.8 Flash is 45% below Google's rate.

On this page
Updated 10 November 2026. SayGM serves every Gemini text model at least 38% below Google's standard rate, and 45% below on 3.8 Flash: Gemini 3.8 Flash is $0.41 per million input tokens and $2.05 output, against Google's $0.75 and $3.75, and Google's new flagship, Gemini 4 Argon, is $1.24 and $6.18 against $2.00 and $10.00. Both of Google's rates are introductory: Argon's doubles on 1 December 2026 and 3.8 Flash's on 1 January 2027.
Gemini API pricing at standard speed, checked 10 November 2026
| Model | Google direct | SayGM | Requesty (+5%) | OpenRouter (+5.5% fee) | Checked |
|---|---|---|---|---|---|
| Gemini 4 Argon | $2.00 / $10.00 · $360 | $1.24 / $6.18 · $222 | $2.10 / $10.50 · $378 | $2.11 / $10.55 · $380 | 10 Nov 2026 |
| Gemini 3.8 Flash | $0.75 / $3.75 · $135 | $0.41 / $2.05 · $74 | $0.79 / $3.94 · $142 | $0.79 / $3.96 · $142 | 10 Nov 2026 |
| Gemini 3.6 Flash | $0.75 / $3.75 · $135 | $0.46 / $2.32 · $83 | $0.79 / $3.94 · $142 | $0.79 / $3.96 · $142 | 10 Nov 2026 |
| Gemini 3.1 Pro (preview) | $2.00 / $12.00 · $400 | $1.24 / $7.42 · $247 | $2.10 / $12.60 · $420 | $2.11 / $12.66 · $422 | 10 Nov 2026 |
| Gemini 3.5 Flash-Lite | $0.30 / $2.50 · $74 | $0.19 / $1.55 · $46 | $0.32 / $2.62 · $78 | $0.32 / $2.64 · $78 | 10 Nov 2026 |
| Gemini 3.1 Flash-Lite | $0.25 / $1.50 · $50 | $0.15 / $0.93 · $31 | $0.26 / $1.58 · $52 | $0.26 / $1.58 · $53 | 10 Nov 2026 |
Each cell shows input / output per million tokens, then the cost of 100 million tokens at an 80/20 input/output mix. Google's rates are its standard paid tier from Google's pricing page; SayGM's are from its live catalog at api.saygm.com/v1/models. Requesty adds its 5% markup and OpenRouter its 5.5% fee to Google's rate. Gemini 3.1 Pro's rates are for prompts up to 200K tokens.
Table of Contents
- How much does the Gemini API cost?
- Gemini 4 Argon pricing
- Gemini 3.8 Flash pricing and the January change
- Gemini 3.1 Pro pricing above 200K tokens
- Gemini Flash-Lite pricing: 3.5 and 3.1
- Where to buy Gemini: Google, OpenRouter, Requesty and SayGM
- How private is Gemini on SayGM?
- What changed since October
- Questions about Gemini API costs
- What to budget for Gemini
- About SayGM
Gemini API pricing now runs from $0.25 per million input tokens for 3.1 Flash-Lite to $2.00 for Gemini 4 Argon and 3.1 Pro, and SayGM serves every one of them below Google's standard rate, through the Gemini SDK you already use. The timing matters more than usual this quarter: Google's rates for Argon double on 1 December, and for 3.8 and 3.6 Flash on 1 January, so a budget built on today's prices will be wrong within weeks.
How much does the Gemini API cost?
SayGM is the cheapest way to buy Gemini at standard speed. SayGM is at least 38% below Google's standard rate on every Gemini model, and 45% below on 3.8 Flash, and Requesty and OpenRouter both add a fee on top of Google's price.
On an 80/20 input-to-output mix, 100 million Gemini 3.8 Flash tokens costs $74 on SayGM, $135 direct from Google and $142 through Requesty or OpenRouter. The same volume of Gemini 4 Argon is $222 on SayGM against $360 direct, and $720 direct once Google's rate doubles on 1 December.

Gemini 4 Argon pricing
Gemini 4 Argon is Google's new flagship reasoning model, built, in Google's words, for complex software engineering, deep mathematical and scientific problem solving, autonomous multi-step agents and long-decode generation up to 1M tokens.
Three things to know before budgeting for it:
The price doubles on 1 December. Google charges $2.00 per million input tokens and $10.00 output through 30 November 2026, then $4.00 and $20.00 from 1 December. Cached input goes from $0.10 to $0.20. A month of Argon at 100 million tokens goes from $360 to $720 direct.
There is no free tier. Argon is paid-only, and Google bills every token the model processes, including its thinking tokens and the internal tool instructions and function declarations it uses for grounding.
Caching is implicit only. Repeated prompt prefixes are cached and billed at the cache rate automatically; you can't set up explicit caches as you can on other Gemini models.
On SayGM, Argon is $1.24 and $6.18 per million, 38% below Google's current rate, or about $222 per 100 million tokens. Until 1 December, Argon also costs less than Gemini 3.1 Pro on output on both platforms, so for most reasoning work it's the better-priced choice as well as the newer model. From 1 December, Google's Argon rate of $4.00 and $20.00 makes it dearer than 3.1 Pro. The Gemini 4 Argon model page shows the live rate.
Gemini 3.8 Flash pricing and the January change
Gemini 3.8 Flash is Google's newest Flash model, which Google calls its most intelligent Flash model, built for long-running coding work and agents. It has a 1M-token context window and returns up to 64K tokens per response.
Four pricing details matter more than the headline rate.
The rate is introductory. $0.75 and $3.75 applies through 31 December 2026. From 1 January 2027 Google's standard rate becomes $1.50 and $7.50, and the same change applies to 3.6 Flash. Cached input doubles too, from $0.075 to $0.15.
Thinking tokens bill as output. 3.8 Flash reasons before it answers, at low, medium or high effort, and Google charges those reasoning tokens at the output rate. A request that thinks hard can bill several times the visible answer's length. Medium is the default; if a task doesn't need it, low effort is the quickest way to cut a 3.8 Flash bill.
Google sells four speeds. Standard is the rate everyone quotes. Flex and Batch are half that, $0.375 and $1.875, for work that can wait. Priority is $1.35 and $6.75 for traffic that can't. All four double on 1 January.
Gemini 2.5 Pro traffic moves to it on 1 December. Gemini 2.5 Pro is deprecated, and from 1 December 2026 Google will automatically route all API requests to gemini-2.5-pro to 3.8 Flash and bill them at 3.8 Flash rates. If you still call 2.5 Pro, your bill changes that day.
SayGM's 3.8 Flash rate, $0.41 and $2.05, is 45% below Google's standard rate and close to Google's own Flex price, at standard speed. The Gemini 3.8 Flash model page shows the live rate and measured throughput.

Gemini 3.1 Pro pricing above 200K tokens
Gemini 3.1 Pro is still in preview, and it has no free tier. Google charges $2.00 input and $12.00 output for prompts up to 200K tokens. Past that, every token in the request costs more: $4.00 input and $18.00 output.
That threshold is easy to cross by accident. A long codebase, a stack of PDFs or an agent carrying a growing conversation history can push a single request past 200K, and the whole request then bills at the higher rate.
SayGM serves 3.1 Pro at $1.24 and $7.42, 38% below Google's standard rate. Until 1 December, Gemini 4 Argon is priced below 3.1 Pro on output at both Google and SayGM, so it's the model to test first for new reasoning work; after Google's rate doubles, 3.1 Pro is the cheaper of the two.
Gemini Flash-Lite pricing: 3.5 and 3.1
Flash-Lite is the model for high-volume work that doesn't need reasoning depth: classification, extraction, routing, short summaries.
- Gemini 3.1 Flash-Lite is the cheapest Gemini model at $0.25 input and $1.50 output on Google, and $0.15 and $0.93 on SayGM, about $31 per 100 million tokens.
- Gemini 3.5 Flash-Lite is newer and costs a little more: $0.30 and $2.50 on Google, $0.19 and $1.55 on SayGM.
Neither is affected by the December or January changes.
Where to buy Gemini: Google, OpenRouter, Requesty and SayGM
Gemini is a closed model, so every route ends at Google. The options differ in what you pay on top and who else is in the path.
SayGM

At least 38% below Google's standard rate on every Gemini model, and 45% below on 3.8 Flash, from $0.15 input for 3.1 Flash-Lite to $1.24 for 4 Argon, with no subscription and no fee on buying credit. It also carries Claude, GPT and open-weight models behind the same key, and its gateway runs inside hardware-attested confidential compute.
Switching is a base URL change in the Google GenAI SDK; the model name stays the same:
import os
from google import genai
from google.genai import types
client = genai.Client(
api_key=os.environ["GM_API_KEY"],
http_options=types.HttpOptions(base_url="https://api.saygm.com"),
)
response = client.models.generate_content(
model="gemini-3.8-flash",
contents="Summarise this contract in five bullet points.",
)
print(response.text)The Gemini compatibility guide covers streaming and REST.
Google AI Studio and Vertex AI

Buying direct gets you every tier: Standard, Flex, Batch and Priority, plus a free tier on the Flash models. The free tier is rate-limited, and Google's pricing page notes that free-tier usage is used to improve its products, which the paid tier is not. Gemini 4 Argon and 3.1 Pro have no free tier. Vertex AI is the same models under Google Cloud billing.
OpenRouter

OpenRouter passes Google's rates through and charges a fee when you buy credit: 5.5% on its Standard plan and 8% on Business, which we break down in our OpenRouter alternatives comparison. On Standard that makes 100 million 3.8 Flash tokens about $142.
Requesty

Requesty passes Google's rate through plus 5%, so 3.8 Flash comes to about $0.79 and $3.94 per million, or $142 per 100 million tokens.
How private is Gemini on SayGM?
Gemini is a frontier model, so Google still receives the prompt; that's true on every route. What SayGM changes is who sees it on the way. Requests use anonymous routing: Google gets the request but never sees who sent it, and the gateway in between runs inside hardware-attested confidential compute.
SayGM can also strip sensitive data out of a prompt before Google ever receives it. Guardrails are optional and set per API key in the dashboard. Once switched on, they redact personal data, secrets and credentials, match your own regular expressions and filter blocked words. They run inside the gateway's enclave, so the redaction happens before the request leaves for Google. They are pattern-based, which means they will miss some sensitive values and catch some harmless ones, so treat them as one layer of protection rather than a complete data-loss-prevention setup.
If you need a prompt nobody can read, including the model's operator, that guarantee comes from SayGM's confidential open-weight models, which run inside the enclave itself.
What changed since October
This post first went up on 5 October 2026. Here's what has moved since.
- Google launched Gemini 4 Argon, its new flagship reasoning model, at $2.00 and $10.00, rising to $4.00 and $20.00 on 1 December 2026. SayGM serves it at $1.24 and $6.18.
- SayGM's Gemini rates moved to at least 38% below Google's standard rate on every model, and 45% below on 3.8 Flash. Gemini 3.8 Flash rose from $0.37 and $1.86 to $0.41 and $2.05; 3.1 Pro fell from $1.50 and $9.00 to $1.24 and $7.42; 3.1 Flash-Lite fell from $0.19 and $1.13 to $0.15 and $0.93.
- Gemini 2.5 Pro requests move to 3.8 Flash on 1 December, billed at 3.8 Flash rates.
- Gemini 3.7 Flash and 3.5 Flash are no longer on Google's pricing page or in SayGM's catalog, so they've left the table.
Questions about Gemini API costs
How much does Gemini 4 Argon cost?
As of 10 November 2026, Gemini 4 Argon costs $2.00 per million input tokens and $10.00 output on Google, rising to $4.00 and $20.00 on 1 December 2026. On SayGM it's $1.24 and $6.18, about $222 per 100 million tokens at an 80/20 mix. It has no free tier.
How much does Gemini 3.8 Flash cost?
Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 output on Google's standard tier until 31 December 2026, then $1.50 and $7.50. On SayGM it's $0.41 and $2.05 as of 10 November 2026, 45% below Google's rate. Thinking tokens bill as output.
Is the Gemini API free?
Partly. Google offers a free tier on the Flash and Flash-Lite models, with rate limits, and notes that free-tier usage is used to improve its products. Gemini 4 Argon and 3.1 Pro have no free tier. For production traffic, the paid tier is the realistic option.
What does Gemini cost per 1,000 tokens?
Divide the per-million rate by 1,000. Gemini 3.8 Flash on Google's standard tier is $0.00075 per 1,000 input tokens and $0.00375 per 1,000 output tokens until January. On SayGM it's about $0.00041 and $0.00205.
Will SayGM's Gemini prices change in December and January?
SayGM's rates track what providers charge, so expect them to move when Google's do. SayGM's model pages always show the current figure.
Which Gemini model should I use?
3.8 Flash for most work: it's the newest Flash model and, until January, among the cheapest. Gemini 4 Argon for the hardest reasoning, coding and agent work; until 1 December it's also priced below 3.1 Pro. Flash-Lite for high-volume tasks where speed and price matter more than depth.
What to budget for Gemini
If you're budgeting Gemini into 2027, use the new rates. Gemini 4 Argon doubles on 1 December and 3.8 Flash on 1 January. On SayGM today, 3.8 Flash is $0.41 and $2.05, Argon is $1.24 and $6.18, and Flash-Lite starts at $0.15.
Set your Gemini client's base URL to api.saygm.com, keep the model name, and say gm to AI.
About SayGM
SayGM is a drop-in inference gateway for teams who don't want to just take a company's word that their prompts are private. Route to a confidential model and your prompt is sealed inside hardware-verified execution, where not even SayGM can see it. That's not a policy, it's provable. Swap in your existing OpenAI, Anthropic, or Gemini code and you're covered in minutes, at transparent, published rates with no hidden markup.
Say gm to AI at saygm.com.
- AI Inference
- Gemini
- Pricing


