NVIDIA models, priced in toman
16 NVIDIA models are live on uttapen. Prices are in toman per 1 million tokens, and you are charged for the tokens you actually use.
NVIDIA on uttapen
NVIDIA currently has 16 models in the uttapen catalogue. 9 of them are free; the rest bill against a toman wallet.
The spread inside NVIDIA alone is wide: Nemotron 3.5 Lightning runs about 122 toman for a 1,000-word request, while Nemotron 3 Ultra costs 1,350 toman for the same one. Picking the right model inside a brand matters as much as picking the brand.
Measured against the whole catalogue, the median NVIDIA output price undercuts about 75% of the other paid models.
On capabilities: 8 take image input, 9 support tool calling, 11 do multi-step reasoning, 1 accepts audio. The longest context belongs to Nemotron 3 Super, at 1,048,576 tokens.
Every NVIDIA model answers on the same sk-up- key and the same base_url: switching between them means changing the `model` string and nothing else.
Which NVIDIA model should I start with?
- Nemotron 3.5 Lightning122 · per 1,000 wordsThe cheapest way in: about 122 toman for a 1,000-word request.
- Nemotron 3 Nano Omni (free)0 · per 1,000 wordsThe cheapest NVIDIA model with tool calling — for agents and for wiring the model to your own API.
- Nemotron 3 Super1,215 · per 1,000 wordsFor multi-step problems, maths and debugging: it reasons before it answers.
- Nemotron 3.5 Content Safety180 · per 1,000 wordsWhen the input is an image or a scanned invoice.
Every NVIDIA model, priced in toman
| Model | Capabilities | Context | Input / 1M | Output / 1M |
|---|---|---|---|---|
Nemotron 3 Nano 30B A3Bnvidia/nemotron-3-nano-30b-a3b | Tool callingReasoningJSON output | 262,144 | 20,773 | 83,093 |
Nemotron 3 Supernvidia/nemotron-3-super-120b-a12b | Tool callingReasoningJSON output | 1,048,576 | 103,866 | 830,926 |
Nemotron 3 Ultranvidia/nemotron-3-ultra-550b-a55b | Tool callingReasoningJSON output | 262,144 | 207,732 | 830,926 |
Nemotron 3.5 Content Safetynvidia/nemotron-3.5-content-safety | Image inputReasoning | 131,072 | 69,244 | 69,244 |
Nemotron 3.5 Lightningnvidia/nemotron-3.5-lightning | Tool callingReasoningJSON output | 262,144 | 24,235 | 69,244 |
Nemotron 3.5 Super VLnvidia/nemotron-3.5-super-vl | Image inputTool callingReasoningJSON output | 1,048,576 | 51,933 | 173,110 |
Switchyardnvidia/switchyard | 1,000,000 | Variable price | ||
Nemotron 3 Nano Omninvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free | Free · shared capacityImage inputTool callingReasoningAudio input | 256,000 | 0 | 0 |
Nemotron 3.5 Embed 1Bnvidia/nemotron-3.5-embed-1b:free | Free · shared capacityImage inputEmbeddings | 32,768 | 0 | — |
Nemotron 3.5 Embed 8Bnvidia/nemotron-3.5-embed-8b:free | Free · shared capacityImage inputEmbeddings | 32,768 | 0 | — |
Nemotron 3 Embed 1Bnvidia/nemotron-3-embed-1b:free | Free · shared capacityEmbeddings | 32,768 | 0 | — |
Llama Nemotron Embed VL 1B V2nvidia/llama-nemotron-embed-vl-1b-v2:free | Free · shared capacityImage inputEmbeddings | 131,072 | 0 | — |
Nemotron 3.5 Super VLnvidia/nemotron-3.5-super-vl:free | Free · shared capacityImage inputTool callingReasoningJSON output | 1,000,000 | 0 | 0 |
Nemotron 3.5 Lightningnvidia/nemotron-3.5-lightning:free | Free · shared capacityTool callingReasoning | 1,000,000 | 0 | 0 |
Nemotron 3.5 Content Safetynvidia/nemotron-3.5-content-safety:free | Free · shared capacityImage inputReasoning | 128,000 | 0 | 0 |
Nemotron 3 Ultranvidia/nemotron-3-ultra-550b-a55b:free | Free · shared capacityTool callingReasoning | 1,000,000 | 0 | 0 |
Common questions about the NVIDIA API
- How do I reach the NVIDIA API from Iran?
- Sign up with your mobile number, top the wallet up in toman and create an sk-up- key. Then, in the official OpenAI SDK, point base_url at https:/
/ api.uttapen.ir/ v1 and set model to the NVIDIA model id. No international card, no VPN. - What do NVIDIA models cost in toman?
- It depends on the model: the cheapest, Nemotron 3.5 Lightning, is about 122 toman for a 1,000-word request, and the dearest, Nemotron 3 Ultra, about 1,350 toman. Each model page carries its own exact price, and you are billed for what the request actually used.
- How many NVIDIA models are on uttapen?
- 16, 9 of them free. 9 support tool calling and 8 accept image input. 4 are embedding models for /v1/
embeddings. The catalogue re-syncs hourly, so a newly released model shows up here on its own. - What are the limits on the free NVIDIA models?
- They have a daily per-user cap and share their capacity with everyone else. When that shared pool runs out the request is refused with a clear error and your wallet is left alone.