Get started
Models
Every model Paranine serves, its id, where it runs and its price per million tokens in rupees.
The catalogue
Send the model id exactly as shown, punctuation included. Prices are in rupees per million tokens.
| Model | Route id (the model string) | Pool | Status | Input, Rs / 1M | Cached input, Rs / 1M | Output, Rs / 1M | Context |
|---|---|---|---|---|---|---|---|
| gpt-oss 120B | paranine/gpt-oss-120b(Global) | Global | Live | Rs 30.08 | Rs 30.08 | Rs 150.39 | 131,072 tokens |
| Kimi K3 | paranine/Kimi-K3(Global) | Global | Live | Rs 902.34 | Rs 90.23 | Rs 4,511.70 | 1,048,576 tokens |
| GLM-5.3 | paranine/GLM-5.3(Global) | Global | Live | Rs 421.09 | Rs 42.11 | Rs 1,323.43 | Not published |
| GLM-5.3 Fast | paranine/GLM-5.3-Fast(Global) | Global | Live | Rs 631.64 | Rs 63.16 | Rs 1,985.15 | Not published |
| Kimi K2.7 Code | paranine/Kimi-K2.7-Code(Global) (proposed, not registered yet) | Global | Coming soon | Rs 285.74 | Same as input | Rs 1,203.12 | Not published |
| DeepSeek V4.1 Flash | paranine/DeepSeek-V4.1-Flash(Global) | Global | Live | Rs 90.23 | Rs 2.11 | Rs 360.94 | Not published |
| DeepSeek-V4-Flash-0731 | paranine/DeepSeek-V4-Flash-0731(Global) (proposed, not registered yet) | Global | Coming soon | Rs 39.10 | Same as input | Rs 78.20 | Not published |
| DeepSeek V4 Pro 0813 | paranine/DeepSeek-V4-Pro-0813(Global) (proposed, not registered yet) | Global | Coming soon | Rs 397.03 | Same as input | Rs 1,191.09 | Not published |
| DeepSeek V4 Pro | paranine/DeepSeek-V4-Pro(Global) (proposed, not registered yet) | Global | Coming soon | Rs 523.36 | Same as input | Rs 1,046.72 | Not published |
| Nvidia Nemotron 3 Ultra | paranine/Nemotron-3-Ultra(Global) (proposed, not registered yet) | Global | Coming soon | Rs 180.47 | Same as input | Rs 721.87 | Not published |
| Inkling | paranine/Inkling(Global) (proposed, not registered yet) | Global | Coming soon | Rs 300.78 | Same as input | Rs 1,218.16 | Not published |
| Inkling-Small | paranine/Inkling-Small(Global) (proposed, not registered yet) | Global | Coming soon | Rs 150.39 | Same as input | Rs 360.94 | Not published |
| gpt-oss 120B | paranine/gpt-oss-120b(high) | PK NPU | Live | Rs 30.08 | Rs 30.08 | Rs 150.39 | 131,072 tokens |
| Qwen3.8 27B | paranine/qwen-3.8-27b | PK NPU | Live | Rs 80.00 | Same as input | Rs 650.00 | 262,144 tokens |
| DeepSeek-OCR 2 | paranine/DeepSeek-OCR2 | PK NPU | Live | Rs 9.02 | Same as input | Rs 9.02 | Not published |
| Laya Decision | paranine/laya-decision | PK NPU | Live | Rs 9.94 | Same as input | No output tokens | Not published |
| Qwen3 Embedding 8B | paranine/qwen3-embedding | PK NPU | Live | Rs 2.78 | Same as input | No output tokens | 32,768 tokens |
A model marked Coming soon is listed ahead of launch: until it opens, requests to it answer 404 (model_not_found). Its id may still change before then.
Where models run
| Pool | Where the model runs |
|---|---|
| PK NPU | Inside Pakistan. Prompts and completions stay in the country. |
| Global | With Paranine's partners outside Pakistan, with Paranine's keys, limits, logs and rupee billing around every request. |
Both pools use the same endpoint, request shape and key. For workloads that must stay in Pakistan, see Data residency.
Choosing a model
- Assistants and tool use: start with gpt-oss 120B, an open-weight reasoning model with a large context window.
- Lowest latency: compare time to first token in the dashboard's analytics, measured on your own prompts.
- Scans and photos of documents: Document OCR.
- Classifying text and yes/no questions: Decisions.
- Search and retrieval: Embeddings.
The dashboard's Model APIs page has each model's details: lab, licence, context window and published benchmarks.
Listing models from code
GET /v1/models is not available yet. The models your organisation can call, with their prices, are returned by GET /api/models on the Account API.
Something unclear or wrong on this page? Tell us. Machine-readable copies: llms.txt.