NanoGPT vs OpenRouter: Same Models, No Deposit Fee
NanoGPT and OpenRouter solve the same problem: one API key, hundreds of models, no juggling separate accounts with every AI lab. Both bill language model usage at the provider's list price.
The difference is what happens to your money before it ever reaches a model - and how much you can verify afterwards.
Pricing: the same list prices, minus the fee
Per-token prices for major models are identical on both platforms, because both charge the model creator's official rates:
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Claude Opus 4.8 | $5.00 | $25.00 |
| Claude Sonnet 4.6 | $3.00 | $15.00 |
| GPT-5.5 | $5.00 | $30.00 |
| Gemini 3.5 Flash | $1.50 | $9.00 |
| DeepSeek V4 Pro | $1.10 | $2.20 |
So where do aggregators make money? On the way in.
OpenRouter charges a fee of roughly 5.5% when you purchase credits (per their published pricing as of June 2026, with a minimum fee around $0.80). Buy $100 of credits and about $5.50 goes to fees before you've sent a single token.
NanoGPT credits deposits in full. Deposit $100, get $100 of balance. There is no percentage fee on card or crypto top-ups. For identical usage, that makes NanoGPT roughly 5% cheaper end to end, despite both platforms advertising the same per-token rates.
Two opt-in NanoGPT surcharges to know about: manually pinning your requests to one specific provider (instead of letting auto-routing pick) adds 5%, and bring-your-own-key requests are billed 5% of the normal model cost. Standard usage - which is what nearly everyone does - has no markup of any kind.
Don't take our word for it
Every chat completion and Responses API response includes the exact amount deducted from your balance for that request. Chat completions return an x_nanogpt_pricing receipt:
"x_nanogpt_pricing": {
"cost": 0.01075,
"currency": "USD",
"inputTokens": 1024,
"outputTokens": 512
}
and the Responses API includes cost and cost_usd directly on the usage object. Multiply your token counts by the model creator's published rates and you'll land on the same number. You can also pull live per-token pricing for every model from https://nano-gpt.com/api/v1/models?detailed=true and audit it against provider price pages whenever you like.
Minimums and payment options
OpenRouter requires buying credits upfront by card or crypto, with the purchase fee applied either way.
NanoGPT starts at $0.10 with crypto or $1 with card. We accept cards, Nano (instant and feeless), Bitcoin, Ethereum, Monero, Solana, USDC, and machine-to-machine payments via the x402 protocol - and you don't need an account to start.
What you get with the API
Both platforms offer an OpenAI-compatible chat completions API with streaming and tool calling. Beyond that, NanoGPT also ships:
- An Anthropic-compatible Messages API (
/v1/messages) for tools built against Claude's native format - An OpenAI Responses API (
/v1/responses) with background mode and response chaining - Image, video, text-to-speech, and transcription models behind the same key and balance - OpenRouter is centered on text
- A full web chat interface included at no extra cost, with conversation history stored locally by default
- Request-level provider routing controls: fallback behavior, price caps, quantization filters, and provider include/exclude lists
Bottom line
If you're paying list price for tokens either way, the deposit fee is the whole ballgame. On OpenRouter, ~5.5% of your spend goes to fees before any model sees a token. On NanoGPT, $100 in means $100 of AI out - verifiable, request by request, in the cost field.
Try it from $0.10 at nano-gpt.com.
