No international card
UPI or netbanking, in ₹. Prepaid only. No surprise recurring debits.
14.56% over mid-market
A 9% markup on usage and the 5.1% card forex fee inside the rate. Top-ups add nothing.
Caching passes through
Requests forwarded unmodified, so cache_control survives the hop completely.
One key, all models
Swap the model string, keep your SDK and retries. Anthropic, OpenAI, Gemini & Qwen.
Two lines. That’s the whole port.
API VERSE speaks the standard OpenAI wire format, so your SDK, retries, and tool-call parsing stay exactly as they are. Change the base URL and the key. Anthropic clients get a native endpoint at /anthropic, so prompt caching works natively.
1 from openai import OpenAI
2
3 client = OpenAI(
4 api_key = os.environ["OPENAI_API_KEY"],
5 )
6
7 resp = client.chat.completions.create(
8 model = "gpt-5.1",
9 messages = msgs,
10 )1 from openai import OpenAI
2
3 client = OpenAI(
4 base_url = "https://apiverse.in/v1",
5 api_key = os.environ["VERSE_API_KEY"],
6 )
7
8 resp = client.chat.completions.create(
9 model = "claude-sonnet-5", # or any frontier ID
10 messages = msgs,
11 )That host is live right now. The service runs exactly as shown.
The providers’ own rates.
The provider’s own list rate, in dollars, per million tokens — check any row against Anthropic’s, OpenAI’s or Google’s published pricing. Converted at the live rate of ₹101.24283 to $1, with 9% added. So a row reading $2.00 costs you ₹220.71 per million.
| # | Model · Provider | Model String | $ / 1M In | $ / 1M Out | $ / 1M Cached |
|---|---|---|---|---|---|
| 01 | Qwen Flash Qwen | qwen/qwen-flash | $0.022 | $0.22 | $0.0043 |
| 02 | Qwen Turbo Qwen | qwen/qwen-turbo | $0.043 | $0.09 | $0.0086 |
| 03 | GPT-5 Nano OpenAI | gpt-5-nano | $0.05 | $0.40 | $0.01 |
| 04 | Doubao Seed 2.0 Mini ByteDance | volcengine/doubao-seed-2.0-mini | $0.06 | $0.56 | $0.02 |
| 05 | Z.ai: GLM-4.7 FlashX Z.ai | glm-4.7-flashx | $0.072 | $0.40 | $0.01 |
| 06 | Google: Gemini 2.5 Flash Lite Google | gemini-2.5-flash-lite | $0.10 | $0.40 | $0.01 |
| 07 | OpenAI: GPT-6 Luna OpenAI | gpt-6-luna | $0.10 | $0.50 | $0.01 |
| 08 | Qwen: Qwen3.5 Flash Qwen | qwen/qwen3.5-flash | $0.10 | $0.40 | $0.01 |
| 09 | Qwen: Qwen3.8 Flash Qwen | qwen/qwen3.8-flash | $0.11 | $0.39 | $0.011 |
| 10 | Qwen Plus Qwen | qwen/qwen-plus | $0.12 | $0.29 | $0.023 |
Signed in to first token in a few minutes.
Sign in with Google
No card, ever. The balance is prepaid, so it cannot overdraft and there is no invoice at the end of the month.
One ClickTop up over UPI
From ₹100. UPI carries 0% MDR by RBI mandate, so ₹100 paid is ₹100 credited — nothing is deducted on the way in.
Credited when PayU confirmsCall any model
Point your existing client at the base URL. Every call lands in Request Logs with the model, tokens, and exact ₹ cost.
2 lines changedYou see the same math we do.
Two things sit between the provider’s price and yours. The markup is 9% on usage. The exchange rate carries a further 5.1% — the fee our card charges to convert rupees into the dollars we fund the upstream account with, passed through at cost rather than absorbed.
Compounded, you pay 14.56% over the mid-market rate. There is no rebate scheme and nothing taken out of a top-up.
The ones worth answering.
Do I need an international card?↓
No. You top up in rupees through PayU — by UPI or netbanking — and we hold the international payment relationship with the upstream providers. That is most of what this is for.
What exactly am I charged for a request?↓
Only the exact token counts returned by upstream provider responses, multiplied by the live converted rate. A request that fails or errors upstream is recorded and charged ₹0.
Does prompt caching survive the hop?↓
Yes. All headers and payload attributes (including Anthropic’s prompt caching breakpoints) pass transparently to provider clusters, and cached rates are applied correctly on return.
Is there a monthly fee or a minimum commitment?↓
None. The account starts at ₹100 prepaid. You can run one single completion or hundreds of thousands. Credits do not expire.
Start with ₹100.
Same key accessing different Models , same rate card, same log — whether you are testing on a Sunday or running production.