Available Models

A list of Neosantara AI Models you can access.

Models

Neosantara AI provides access to various AI models with different capabilities, optimized for diverse needs:

Indonesian-Optimized Models

nusantara-base

• Description: A specialized model for Indonesian content with a deep awareness of cultural context.
• Context Window: 64000 Tokens
• Capabilities: function_calling, json_mode
• Best for: Understanding cultural nuances, local context, and complex Indonesian texts

archipelago-7b

• Description: Model khusus untuk konten Bahasa Indonesia dengan kesadaran konteks budaya
• Context Window: 4096 Tokens
• Capabilities: json_mode
• Best for: Pemahaman budaya dan konteks lokal

bahasa-llm

• Description: A fast and efficient model optimized for Indonesian language understanding and rapid responses.
• Context Window: 8192 Tokens
• Capabilities: function_calling, json_mode
• Best for: Fast processing, concise understanding, and function calls within an Indonesian context.

General Purpose Models

gemini-1.5-flash

• Description: A newer Google Gemini Flash model, optimized for quick answers and efficient responses.
• Context Window: 1000000 Tokens
• Capabilities: text_generation, function_calling, json_mode
• Best for: Factual questions, quick summaries, and general tasks.

gemini-2.0-flash

• Description: A general purpose Google Gemini model for fast questions answers.
• Context Window: 32768 Tokens
• Capabilities: text_generation, function_calling, json_mode
• Best for: General conversations, text generation, and automation tasks.

gemini-1.5-pro

• Description: A powerful Google Gemini Pro model for complex tasks and multimodal processing (text and images).
• Context Window: 1000000 Tokens
• Capabilities: function_calling, json_mode
• Best for: In-depth analysis, complex reasoning, and creative tasks

gpt-3.5-turbo

• Description: A general-purpose model from OpenAI, fast and responsive for various use cases.
• Context Window: 16385 Tokens
• Capabilities: text_generation, function_calling, json_mode
• Best for: General conversations, text generation, and automation tasks.

gpt-4o-mini

• Description: A compact model from OpenAI, efficient and capable of handling complex tasks at an affordable price.
• Context Window: 128000 Tokens
• Capabilities: text_generation, function_calling, json_mode
• Best for: High-accuracy tasks and cost-efficiency.

command-r

• Description: Cohere’s Command-R model, offering solid reasoning and language understanding capabilities.
• Context Window: 128000 Tokens
• Capabilities: text_generation, json_mode
• Best for: Natural language understanding and summarization tasks.

command-r-plus

• Description: Cohere’s Command-R Plus model, featuring advanced reasoning and web search capabilities.
• Context Window: 128000 Tokens
• Capabilities: text_generation, web_search, json_mode
• Best for: Questions requiring real-time information or external references.

firefunction-v1

• Description: Fireworks model with function calling capabilities.
• Context Window: 32768 Tokens
• Capabilities: text_generation, function_calling, json_mode
• Best for: Text generation requiring custom function integrations.

replicate-prediction

• Description: A general-purpose Llama2 model hosted on Replicate.
• Context Window: 4096 Tokens
• Capabilities: text_generation, json_mode
• Best for: General text generation tasks with flexible deployment.

gemma2-9b-it

• Description: A Gemma2 model from Groq, optimized for speed and efficiency in language processing.
• Context Window: 8192 Tokens
• Capabilities: text_generation, json_mode
• Best for: Fast responses and lightweight text generation.

llama-3.3-70b-instruct

• Description: Llama 3.3 70B optimized to be faster.
• Context Window: 24000 Tokens
• Capabilities: text_generation, function_calling
• Best for: High-performance text generation and complex instructions.

luminai

• Description: LuminAI gateway to various models. Target model can be changed via options.
• Context Window: 8192 Tokens
• Capabilities: text_generation
• Best for: Flexible model access via LuminAI’s integrated gateway.

vision-emas-2045

• Description: Advanced vision model for analyzing ID cards, receipts, and other visual tasks, optimized for Indonesian context.
• Context Window: 1000000 Tokens
• Capabilities: vision, json_mode, text_generation
• Best for: Image analysis and visual data extraction.

Image Generation Models

neosantara-gen-2045

• Description: Neosantara AI main image generation model, hosted on Together AI.
• Context Window: 0 Tokens
• Capabilities: image_generation
• Best for: Creating high-quality images from text prompts.

replicate-stable-diffusion-xl

• Description: Stable Diffusion XL on Replicate as fallback for image generation.
• Context Window: 0 Tokens
• Capabilities: image_generation
• Best for: Generating diverse images as a reliable alternative.

Embedding Models

nusa-embedding-0001

• Description: Text embedding model for search and recommendation systems.
• Context Window: 0 Tokens
• Capabilities: embedding
• Best for: Converting text into numerical vectors for similarity tasks.

together-embed-v2-L

• Description: Together AI Embeddings v2 Large for high-quality text embeddings.
• Context Window: 0 Tokens
• Capabilities: embedding
• Best for: Advanced semantic search and large-scale text similarity.

Model Selection Guide

Use this guide to choose the right model for your needs:

1. Indonesian Content & Local Context:

   • Prioritize: nusantara-base, archipelago-7b, bahasa-llm

2. Fast & Efficient Processing:

   • Prioritize: bahasa-llm, gemini-1.5-flash, gemini-2.0-flash, gemma2-9b-it

3. Information Retrieval & In-depth Analysis:

   • Prioritize: command-r-plus (for web search), gemini-1.5-pro (for analysis)

4. Function Calling:

   • Prioritize: nusantara-base, bahasa-llm, gemini-1.5-flash, gemini-1.5-pro, gpt-3.5-turbo, gpt-4o-mini, firefunction-v1, llama-3.3-70b-instruct

5. Image Generation:

   • Use: neosantara-gen-2045, replicate-stable-diffusion-xl

6. Image Analysis (Vision):

   • Use: vision-emas-2045

7. Embedding (Text Mapping):

   • Use: nusa-embedding-0001, together-embed-v2-L

Model Capabilities

Here is a list of the main capabilities supported by Neosantara AI models:

Capability	Description
function_calling	The ability of the model to call external functions
json_mode	The ability of the model to generate output in structured JSON format
web_search	The ability of the model to search for information from the web
streaming	Supports streaming responses (gradual output)
image_generation	The ability of the model to create images from text prompts
vision	The ability of the model to analyze and understand content from images
embedding	The ability of the model to convert text into numerical vector representations