Models & Credits
Choose a model
for the task
Use included models within your allowance,
or spend credits on models that charge for usage.
The models
Two ways to pay. Or not.
Included models are free within 50,000 input and output tokens a month, counted across everything you send and receive. The allowance resets one month after you opened the account (UTC, on that same day of the month) and does not roll over. The rest run on credits: one credit is one US cent.
Images go to models that read them; PDF and text files are read as extracted text by all of them.
- Included
Qwen 3.5
General-purpose model with reasoning.
Text, images, PDF and text files · Within the monthly allowance
Ministral 3
Compact, fast text model.
Text, PDF and text files · Within the monthly allowance
Kimi K2.6
General-purpose model with reasoning.
Text, images, PDF and text files · Within the monthly allowance
GLM 5.3 Flash
Runs in a hardware-protected environment through NEAR AI. Diogenes checks cryptographic evidence for the inference service. Diogenes still processes your messages to provide chats, JARs and browsing.
Text, PDF and text files · Within the monthly allowance
- Uses credits
Qwen 3.8 Max
The largest Qwen model offered, with reasoning.
Text, images, PDF and text files · 200 credits / 1M in · 600 credits / 1M out
DeepSeek V4 Pro
Reasoning model, strong on code.
Text, PDF and text files · 87 credits / 1M in · 174 credits / 1M out
Kimi K3
General-purpose model with reasoning.
Text, images, PDF and text files · 255 credits / 1M in · 1295 credits / 1M out
How credits work
Set aside before, counted after
Per token, per reply, from what the model actually read and wrote. Nothing is charged automatically.
- What is charged
- Only replies funded by credits: flagship models, and included models only after you confirm paid continuation in that chat. Sending, regenerating or editing in such a chat needs a balance above zero to start. Included usage within the allowance never touches the balance.
- Input and output
- Charged separately, per million tokens, each side rounded up to a whole credit. Input is everything the model reads: the conversation so far, text from attached files, and in a JAR its instructions and Knowledge. Fetched pages and each step of a browsing task count as input. Long chats cost more per reply.
- Before and after the reply
- Before a reply starts, the amount it may need is set aside from your balance or allowance. When the reply finishes, the actual usage is charged once and the rest is released. A reply that would not fit the remaining balance or allowance is refused before it starts instead of being cut short.
- When it runs out
- Included models pause until the first of next month (UTC); the date is shown in the chat and in Settings. They continue on credits only after you confirm in that chat. The confirmation lasts until the allowance resets or you switch it off, and is asked again if the model or its price changes. Credit models are unaffected by the allowance. Credits do not expire.
- Spending limits
- An optional monthly cap on paid spending is set in Settings → Credits and enforced before each reply. When the service's own free capacity is exhausted, free use pauses with a clear message and nothing is charged.
Top up
USDC in, credits out
Signing in requires no transaction or gas fee. Topping up is a separate wallet transaction.
- Asset
- USDC
- Network
- Ethereum mainnet
- Amount
- 100–100,000 credits ($1.00–$1,000.00)
- Rate
- 1 credit = $0.01
- Credited after
- 12 confirmations
The wallet signs two transactions: an approval, then the payment through a verified payment contract that forwards the full USDC amount to Diogenes with a reference for your top-up. Network fees are paid in ETH by your wallet and are not converted to credits; the USDC amount is credited in full.
Credits appear once the payment has twelve confirmations — usually a few minutes. Keep the Credits page open while it checks; the balance updates as soon as the payment is verified.
JARs
Shared Knowledge, own bill
In a JAR, the instructions and shared Knowledge travel with every request, so they count as input.
The person whose chat it is pays: on an included model the tokens count toward their allowance, on a credit model the reply is charged to their balance. There is no separate JAR fee. More in the FAQ.
Your balance and usage are in Settings → Credits.