Billing & Pricing: For a detailed explanation of how credits work, pricing mechanics (Pay-As-You-Go), and throughput tiers, please refer to the Token Credits & Pricing documentation.
Featured Models
Nusantara-Base
Our flagship model. Balanced performance for general tasks, optimized for Indonesian context.
- ID:
nusantara-base - Context: 64k
Garda-Beta-Mini
Lightning fast and efficient. Best for high-volume tasks and long context handling.
- ID:
garda-beta-mini - Context: 131k
Text, Reasoning & Coding Models
Below is the list of available text generation, reasoning, and coding models.| Model ID | Description | Context | Pricing (Input/Output) |
|---|---|---|---|
nusantara-base | Flagship. General purpose, vision-capable, balanced. | 64k | Rp 300 / 1,500 |
garda-beta-mini | Fastest. High efficiency, massive context. | 131k | Rp 2,500 / 10,499 |
archipelago-70b | Deep Culture. Llama-3.3 70B fine-tuned for cultural nuance. | 24k | Rp 4,710 / 36,543 |
sea-lion-v4-27b-it | Regional. Specialized for Southeast Asian languages. | 128k | Rp 350 / 560 |
deepseek-chat-v3.1 | Hybrid. 671B MoE model. | 164k | Rp 3,333 / 8,333 |
kimi-k2 | MoE. 1T params (MoE) optimized for agentic tasks. | 131k | Rp 16,642 / 49,929 |
kimi-k2:latest | MoE. The latest standard Kimi K2 model. | 128k | Rp 16,642 / 49,929 |
kimi-k2:search | Search. Kimi K2 with online search capability. | 128k | Rp 16,642 / 49,929 |
kimi-k2:research | Research. Optimized for deep research tasks. | 128k | Rp 16,642 / 49,929 |
kimi-k2:math | Math. Optimized for complex mathematical logic. | 128k | Rp 16,642 / 49,929 |
kimi-k2:silent | Silent. Direct answers without search process logs. | 128k | Rp 16,642 / 49,929 |
kimi-k1 | Legacy. The first generation Kimi model. | 128k | Rp 16,642 / 49,929 |
llama-3.3-nemotron-super-49b-v1.5 | Super. 49B model with exceptional reasoning. | 132k | Rp 4,560 / 31,991 |
glm-4.6 | Z.ai. SOTA coding and agent benchmarks. | 128k | Rp 823 / 2,823 |
glm-4.6-plus | Z.ai. More powerful coding context. | 128k | Rp 823 / 2,832 |
grok-4.1-fast-non-reasoning | Massive Context. xAIās best tool-calling model. | 2M | 0.72 (USD) |
llama-3.3-70b-instruct | Standard. Robust general intelligence. | 24k | Rp 4,560 / 31,991 |
llama-3.3-70b-turbo | Turbo. Optimized version of Llama 70B. | 8k | Rp 300 / 1,200 |
gemma-3-27b-it | New. Googleās latest open model. | 32k | Rp 200 / 1,500 |
gemma2-9b-it | Efficient. High quality for its size. | 8k | Rp 200 / 1,000 |
llama-3.2-11b | Light. Efficient vision-capable model. | 8k | Rp 100 / 400 |
granite-3-8b-instruct | Enterprise. IBMās enterprise-grade model. | 128k | Rp 2,880 / 11,520 |
gpt-oss-20b | Low Latency. Medium-sized open-weight model. | 131k | Rp 400 / 1,600 |
gemini-3-flash | Speed. Googleās most intelligent model built for speed. | 1M | 3.00 (USD) |
qwen3-coder | Coding. 480B MoE specialist. | 262k | Rp 3,333 / 8,333 |
grok-code-fast | Coding. xAIās latest coding model. | 256k | 0.20 / \1.50 (USD) |
claude-4.5-sonnet | Gateway. Access Anthropic via Vercel. | 200k | 3.00 / \15.00 (USD) |
Image & Vision Models
Models specialized for image generation and optical character recognition (OCR).| Model ID | Description | Pricing |
|---|---|---|
neosantara-gen-2045 | Image. High-speed, high-quality generation. | 1024px: Rp 1,500 512px: Rp 800 256px: Rp 400 |
deepseek-ocr | Vision. High-accuracy text extraction. | Rp 100 / image |
Embeddings
Models for generating text embeddings, useful for search, clustering, and RAG applications.| Model ID | Description | Pricing |
|---|---|---|
nusa-embedding-0001 | Embedding. Text embedding for search and RAG. | Rp 100 / 1M |
gemini-embedding-001 | Embedding. Gemini embedding model. | Free |
Pricing Note: āRpā indicates Indonesian Rupiah. āUSDā indicates US Dollars. Prices are per 1 Million tokens unless otherwise stated.