Cookies on AIGROW

We use necessary cookies to run the site. With your permission, we also use cookies for analytics, session recordings and marketing attribution.

Cookie Policy

Claude, GPT, Gemini and Grok API, up to 75% below list price

Claude Opus 5.5: list price $4.00 through AIGROW credit $1.60 per million input tokens

Certified genuine by AIGROW on 11 November 2026: 9 of 9 models on sale matched the same model on its maker’s API.

How each model is checked

A working key minutes after payment. Works in Claude Code, Cline, Aider, OpenCode and 9 more. One-off packs never renew.

$100 of API credit for $25, on one key for Claude, GPT, Gemini and Grok Every plan and pack

Certified genuine by AIGROW

A cut-price API can answer with a cheaper or older model under the name you asked for. Certified genuine by AIGROW means the model passed AIGROW’s own check against its maker’s API on 11 November 2026; it is not an endorsement by Anthropic, OpenAI, xAI or Google. A model goes on sale only after it passes. The check is run again on every model on sale in the first week of each month, before a model moves to a new supplier and before a new model goes on sale, and a model that fails comes off sale the same day.

9 of 9

models on sale matched their makerCertified genuine by AIGROW on 11 November 2026

Right answers out of 12, the same prompts sent to both sides.
ModelAs soldAt the maker
On sale at AIGROW
Claude Fable 5.212 of 1212 of 12
Claude Opus 5.512 of 1212 of 12
Claude Sonnet 5.512 of 1212 of 12
Claude Fable 5.112 of 1212 of 12
GPT-6 Astra12 of 1212 of 12
GPT-6.1 Sol12 of 1212 of 12
GPT-6.1 Terra12 of 1212 of 12
Grok 4.75 of 55 of 5
Gemini 4 Argon12 of 1211 of 12
Discount API sellers we tested
Sold as Claude Opus 5.54 of 1212 of 12
Sold as Claude Sonnet 5.50 of 1212 of 12

GPT-6.1 Terra and Gemini 4 Argon: the better of two runs, the second with a reasoning effort set. The battery for Grok 4.7 had 5 of the 12 items.

12 questions with one right answer
Sent in one message to the line and to the maker. A model has to reason its way to each answer, so a cheaper model sold under the same name gets fewer.
The same tool call
One request with a weather tool. 9 of 9 models filled in the arguments the maker’s model returns.
The prompt tokens the maker counts
Each model counts a prompt its own way, and a hidden instruction adds to the count. 7 of 9 models report exactly the maker’s number; the other 2 add context of their own, which the endpoint takes off the bill.
Signed thinking on Claude
Anthropic signs the thinking block of every Claude answer. 4 of 4 Claude models brought it back signed.

A line that fails stops serving requests: 2 did in this run. In November 2026 we bought credit from 12 discount API sellers and ran the same check on what they sell: 12 of them had at least one line that failed.

Download the script and run it with your keyEvery result beside its maker

Claude, GPT, Gemini and Grok API pricing, never above list

USD per million tokens, each AIGROW rate drawn against the maker’s own list price as of 12 November 2026. The line between the two is what you save.

Kind of token

Certified genuine by AIGROW on 11 November 2026

  1. Claude Opus 5.5

    Certified genuine by AIGROW on 11 November 2026

    claude-opus-5.560% below list on input, output and cache writes; cache reads at list

    Output: $8.00 list $20.00

    60% below list

  2. Claude Fable 5.2

    Certified genuine by AIGROW on 11 November 2026

    claude-fable-5.260% below list on input, output and cache writes; cache reads at list

    Output: $16.00 list $40.00

    60% below list

  3. Claude Fable 5.1

    Certified genuine by AIGROW on 11 November 2026

    claude-fable-5.160% below list on input, output and cache writes; cache reads at list

    Output: $20.00 list $50.00

    60% below list

  4. GPT-6 Astra

    Certified genuine by AIGROW on 11 November 2026

    gpt-6-astra75% below list on every kind of token

    Output: $12.50 list $50.00

    75% below list

  5. Claude Sonnet 5.5

    Certified genuine by AIGROW on 11 November 2026

    claude-sonnet-5.560% below list on input, output and cache writes, 25% on cache reads

    Output: $4.00 list $10.00

    60% below list

  6. GPT-6.1 Sol

    Certified genuine by AIGROW on 11 November 2026

    gpt-6.1-sol75% below list on input, output and cache writes; cache reads at list

    Output: $2.50 list $10.00

    75% below list

  7. GPT-6.1 Terra

    Certified genuine by AIGROW on 11 November 2026

    gpt-6.1-terra75% below list on input, output and cache writes; cache reads at list

    Output: $1.25 list $5.00

    75% below list

  8. Grok 4.7

    Certified genuine by AIGROW on 11 November 2026

    grok-4.775% below list on every kind of token

    Output: $1.50 list $6.00

    75% below list

  9. Gemini 4 Argon

    Certified genuine by AIGROW on 11 November 2026

    gemini-4-argon50% below list on every kind of token

    Output: $5.00 list $10.00

    50% below list

USD per million tokens on a logarithmic scale, where the same saving draws the same length. The ring is the maker’s list price, the dot the rate through AIGROW credit.

Google lists Gemini 4 Argon at $2.00 / $10.00 per million input and output tokens before 1 December 2026, and at $4.00 / $20.00 from that day. The AIGROW rate of $1.00 / $5.00 is 50% below the first price and 75% below the second.

These are the models the endpoint serves, picked as the strongest current ones; nothing else is sold. The rates are AIGROW’s own, never above the provider’s list price, and a request is billed at the rates in force when it is made (Terms of Service, section 7). The model price list compares every other model.

In the figure a cache write costs 1.25 times input, the 5-minute cache. On Claude models every cache write goes to the 1-hour cache, so it costs 2 times input, set against Anthropic’s own price for a 1-hour write.

AI API cost calculator: your month at list price and through credit

Pick a model and how you use it, or type what you pay the maker today. The month is priced twice from the rates above, so the saving you see is your own.

How you use it

250 requests a working day

$a month

Leave it empty for the volume of the usage you picked.

Claude Opus 5.5, Coding agent: $252.00 a month at the list price, $138.60 through AIGROW credit, $113.40 less (45%).

A year $1,360.80 saved: $1,663.20 through AIGROW credit against $3,024.00 at the list price.

This use takes 92% of its credit each month.

Prefer one payment? See the one-off packs

How we count

A developer in Claude Code, Cline or OpenCode, 250 requests on each of 21 working days. Each request reads 60,000 tokens of conversation and files from the prompt cache, writes 2,400 new ones to it, sends 1,200 uncached and gets 600 back.

Cache reads on Claude Opus 5.5 cost the list rate through credit too, so a use that reads mostly from the cache saves less than the headline discount.

Cache writes on Claude Opus 5.5 go to the 1-hour cache, so both columns price them at 2 times input.

Each cost is the millions of tokens times the rate per million: AIGROW’s from the catalogue the endpoint bills with, Anthropic’s from its list as of 12 November 2026.

A month of Coding agent on Claude Opus 5.5, by kind of token
TokensMillionsAnthropic listAIGROW credit
Cache reads315$63.00× $0.20$63.00× $0.20
Cache writes12.6$100.80× $8.00$40.32× $3.20
Input6.3$25.20× $4.00$10.08× $1.60
Output3.15$63.00× $20.00$25.20× $8.00
A month$252.00$138.60

Buy API credits from $25, by the month or in a pack

The same prices either way. A plan renews each month with 10% to 25% more credit and cancels any time; a pack is one payment. The redemption code is emailed the moment payment clears.

Your credit works with

  • Claude Opus 5.5, Claude Fable 5.2, Claude Fable 5.1, Claude Sonnet 5.5
  • GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra
  • Gemini 4 Argon
  • Grok 4.7

Monthly plansCancel any time

  • $110 of API credit every month

    Includes a 10% plan bonus

    $4.40 of credit per $1

  • $230 of API credit every month

    Includes a 15% plan bonus

    $4.60 of credit per $1

  • $600 of API credit every month

    Includes a 20% plan bonus

    $4.80 of credit per $1

  • Best value

    $1,250 of API credit every month

    Includes a 25% plan bonus

    $5.00 of credit per $1

One-off packsNo renewal

  • $100 of API credit

    $4.00 of credit per $1

  • $200 of API credit

    $4.00 of credit per $1

  • $500 of API credit

    $4.00 of credit per $1

  • $1,000 of API credit

    $4.00 of credit per $1

Is it the real model?
Certified genuine by AIGROW on 11 November 2026: 9 of 9 models on sale matched the same model on its maker’s API, and a model that fails comes off sale.
How fast does the key come?
The redemption code is emailed the moment payment clears. Redeem it in API access and create a key: the first request can go out minutes after you pay.
What happens to credit I do not use?
It stays on the account when a plan is cancelled, and stays usable on the other models if one is withdrawn. Spent credit is not refunded; unspent credit you paid for is refunded only as the Refund Policy provides (Terms of Service, section 7).

Claude Code, Cline, Aider or the OpenAI SDK, connected in three steps

The endpoint speaks both the Anthropic and the OpenAI protocol, so each tool keeps its own SDK and only the base URL and the key change.

  1. Buy credit

    Pay by card at the Stripe checkout. The redemption code is emailed the moment payment clears.

  2. Redeem the code and create a key

    Sign in, paste the code into API access and create a key that starts with ag_. The code is not tied to the email you paid with.

  3. Set the base URL

    Point your tool or SDK at the endpoint below and use your key. Requests are billed per token from your credit.

OpenAI SDK, Python

Model
from openai import OpenAI

client = OpenAI(
    base_url="https://aigrow.gg/v1",
    api_key="ag_your_key_here",
)

resp = client.chat.completions.create(
    model="claude-opus-5.5",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)
$1.60 / $8.00 per million input / output tokens through creditlist $4.00 / $20.00

  • Anthropic protocol

    For Claude Code, Claude Desktop and the Anthropic SDK.

    https://aigrow.gg

  • OpenAI protocol

    For Cline, Aider, OpenCode, the OpenAI SDK and most other tools.

    https://aigrow.gg/v1

Setup guides by tool

The setup guide has the same blocks for the OpenAI and Anthropic SDKs and curl, with the endpoints and the rate limits.

What the terms promise about your bill

Four rules from the Terms of Service that decide what a request can cost.

No rate above list
Every rate through credit is at or below the provider’s list price, cache reads and cache writes included.
No change to a request already made
AIGROW may change its rates or the models it offers. The rates in force are the ones in the dashboard, and a change never applies to a request already made.
No charge for a refused request
A request the endpoint turns away before it reaches the model, because it is busy or the model is not offered, is not billed.
A plan stops when you say
Cancel from the link in any plan email. The plan ends with the month already paid, and the credit already added stays on the account.

Spent credit is not refunded or exchanged for cash. The full rules are in the Terms of Service, section 7.

Compare AI API prices before you buy

Each page works out one question from the same rates and list prices.

Want ready-made tools rather than an API? ToolSuite, a partner of ours, sells 50+ premium AI and ecommerce tools on one subscription. What ToolSuite includes and what it costs

AIGROW sells prepaid API credit for Claude Opus 5.5, Claude Fable 5.2, Claude Fable 5.1, GPT-6 Astra, Claude Sonnet 5.5, GPT-6.1 Sol, GPT-6.1 Terra, Grok 4.7 and Gemini 4 Argon through one endpoint, each checked against the same model at its maker. Credit costs a quarter of its face value: $25 buys $100. Rates are published per token, 50% to 75% below list on input and output, never above list. Monthly plans add 10% to 25% bonus credit.

AI API credit questions: models, price and setup

Whether the model is genuine, how the discount is worked out, and what arrives when.

Yes. AIGROW credit costs a quarter of its face value, so Claude Opus 5.5 works out at $1.60 per million input tokens and $8 per million output tokens, against a list price of $4 and $20. The same key also runs GPT, Gemini and Grok models.

Each model on sale is run beside the same model at its maker, and you can repeat the check with your own key. Both sides get up to 12 questions with one right answer, a tool call and fixed prompts whose tokens are counted, and a Claude answer has to bring back the thinking block signed as Anthropic signs it. On 11 November 2026, 9 of 9 models on sale passed, and the 2 lines that failed were taken out of service.

A genuine model names its maker but usually cannot state its exact version, and the makers’ own endpoints answer the same way. So the question cannot tell a genuine model from a substitute, and neither can the model name in the response, which the seller writes. The check compares answers to questions with one right answer and the prompt tokens counted instead.

As of 12 November 2026, per million input / output tokens: Claude Opus 5.5 $1.60 / $8.00 through AIGROW credit (list $4.00 / $20.00); Claude Fable 5.2 $3.20 / $16.00 through AIGROW credit (list $8.00 / $40.00); Claude Fable 5.1 $4.00 / $20.00 through AIGROW credit (list $10.00 / $50.00); GPT-6 Astra $2.50 / $12.50 through AIGROW credit (list $10.00 / $50.00); Claude Sonnet 5.5 $0.80 / $4.00 through AIGROW credit (list $2.00 / $10.00); GPT-6.1 Sol $0.50 / $2.50 through AIGROW credit (list $2.00 / $10.00); GPT-6.1 Terra $0.25 / $1.25 through AIGROW credit (list $1.00 / $5.00); Grok 4.7 $0.50 / $1.50 through AIGROW credit (list $2.00 / $6.00); Gemini 4 Argon $1.00 / $5.00 through AIGROW credit (list $2.00 / $10.00).

You buy credit at a quarter of its face value. Each request takes credit at its model’s rate, so you pay a quarter of that rate: up to 75% below the provider’s list price, with each model’s saving shown beside its name. Prompt-cache reads cost the list price or less, and no rate is ever above list. The dashboard shows rates in credit, four times the figures in the table.

Counted the way the cost calculator on this page counts it, 250 requests a working day on Claude Opus 5.5 come to $138.60 a month through credit, against $252 at the list price: 45% less. A coding agent reads most of each request from the prompt cache, so the saving depends on what the model charges for cache reads. Type your own monthly spend into the calculator to scale it.

Per token, like input and output. A cache write costs 1.25 times the model’s input rate for the 5-minute cache and 2 times for the 1-hour cache, and on Claude models prompts are written to the 1-hour cache. A cache read costs the cache rate in the price table, which on some models is the list rate itself. That is why a coding agent, which reads most of each request from the cache, saves less than the input and output discounts suggest. No rate is ever above list.

The 9 in the table: Claude Opus 5.5, Claude Fable 5.2, Claude Fable 5.1, GPT-6 Astra, Claude Sonnet 5.5, GPT-6.1 Sol, GPT-6.1 Terra, Grok 4.7 and Gemini 4 Argon, current models from Anthropic, OpenAI, Google and xAI. A request for any other model is refused before it is sent, and nothing is billed. Coding agents can also use the provider’s own spelling of each id.

Yes. Claude Code and the Anthropic SDK take the base URL https://aigrow.gg; Cline, Aider, OpenCode and the OpenAI SDK take https://aigrow.gg/v1. In Claude Code, also set ANTHROPIC_DEFAULT_HAIKU_MODEL to claude-sonnet-5.5 so its background calls run on a model the endpoint serves. Codex works only with the Responses API, so it is not one of the supported tools. The setup guide has a page for each supported tool.

https://aigrow.gg for Anthropic-compatible tools and https://aigrow.gg/v1 for OpenAI-compatible ones, with a key that starts with ag_. The model id is the one under each name in the price table, such as claude-opus-5.5; the provider’s own spelling works too. GET /v1/models with your key returns the live list.

At once. Your redemption code is emailed the moment payment clears. Sign in to the dashboard, paste it into API access, and create a key for the AIGROW endpoint. The same page shows your usage and remaining credit.

Only if you choose one. A monthly plan charges the same amount each month and adds 10% to 25% more credit than the pack at that price; the first month comes as a code, later months go straight to the account that redeemed it. Cancel any time from the link in each plan email, or from API access when signed in with the email that paid. A pack is a one-off payment. There is no yearly plan.

Yes, as on every model API. A request the endpoint cannot take right away gets a 429 with a retry-after header, which the OpenAI and Anthropic SDKs follow on their own, and it is not billed.

No. AIGROW is independent and runs its own endpoint. Model and provider names identify what the endpoint serves and do not imply a partnership or endorsement.

Pay up to 75% less for the same models.

$25 for $100 of API creditEach model is checked against the same model at its maker. Start with the $25 pack, run the check script and your own workload on it, and buy more once both hold up.

Compare plans and packs

Certified genuine by AIGROW on 11 November 2026: 9 of 9 models on sale matched the same model on its maker’s API. The redemption code arrives by email the moment payment clears.

$25 for $100 of API credit