Model Catalog

Complete reference of available chat and embedding models with selection guidance.

Chat Models

OpenAI GPT Models

GPT-4o

GitHubModelsChatModelName.GPT_4_O  // "gpt-4o"

Capabilities:

• Latest GPT-4 model
• Multimodal (text + vision)
• Tool calling
• JSON schema responses
• 128K context window

Best For:

• Complex reasoning tasks
• Multi-step problem solving
• Vision/image understanding
• High-quality content generation
• Tasks requiring latest capabilities

Performance: High quality, moderate speed

GPT-4o Mini

GitHubModelsChatModelName.GPT_4_O_MINI  // "gpt-4o-mini"

Capabilities:

• Efficient smaller model
• Tool calling
• JSON responses
• 128K context window

Best For:

• Fast responses
• Simple tasks
• High-volume processing
• Cost-sensitive applications

Performance: Good quality, fast speed

Microsoft Phi Models

Phi-3.5 Mini Instruct

GitHubModelsChatModelName.PHI_3_5_MINI_INSTRUCT  // "Phi-3.5-mini-instruct"

Best For: Resource-constrained scenarios, edge deployment, simple tasks

Phi-3.5 Vision Instruct

GitHubModelsChatModelName.PHI_3_5_VISION_INSTRUCT  // "Phi-3.5-vision-instruct"

Capabilities: Vision understanding

Best For: Image analysis, vision tasks, multimodal applications

Phi-3 Medium/Small/Mini Variants

GitHubModelsChatModelName.PHI_3_MEDIUM_INSTRUCT_128K  // 128K context
GitHubModelsChatModelName.PHI_3_MEDIUM_INSTRUCT_4K    // 4K context
GitHubModelsChatModelName.PHI_3_SMALL_INSTRUCT_128K   // 128K context
GitHubModelsChatModelName.PHI_3_SMALL_INSTRUCT_8K     // 8K context
GitHubModelsChatModelName.PHI_3_MINI_INSTRUCT_128K    // 128K context
GitHubModelsChatModelName.PHI_3_MINI_INSTRUCT_4K      // 4K context

Choose By:

• Long context: 128K variants
• Short interactions: 4K/8K variants
• Size: Mini < Small < Medium (speed vs. capability tradeoff)

AI21 Jamba Models

GitHubModelsChatModelName.AI21_JAMBA_1_5_LARGE  // "ai21-jamba-1.5-large"
GitHubModelsChatModelName.AI21_JAMBA_1_5_MINI   // "ai21-jamba-1.5-mini"
GitHubModelsChatModelName.AI21_JAMBA_INSTRUCT   // "ai21-jamba-instruct"

Best For: Instruction following, structured tasks

Choose:

• Large: Maximum capability
• Mini: Faster, efficient

Cohere Command Models

GitHubModelsChatModelName.COHERE_COMMAND_R        // "cohere-command-r"
GitHubModelsChatModelName.COHERE_COMMAND_R_PLUS   // "cohere-command-r-plus"

Best For: Strong instruction following, command-style interactions, RAG applications

Choose:

• R Plus: Enhanced capabilities
• R: Standard performance

Meta Llama Models

Llama 3.1 (Latest)

GitHubModelsChatModelName.META_LLAMA_3_1_405B_INSTRUCT  // "meta-llama-3.1-405b-instruct"
GitHubModelsChatModelName.META_LLAMA_3_1_70B_INSTRUCT   // "meta-llama-3.1-70b-instruct"
GitHubModelsChatModelName.META_LLAMA_3_1_8B_INSTRUCT    // "meta-llama-3.1-8b-instruct"

Latest Llama generation with improved capabilities

Llama 3

GitHubModelsChatModelName.META_LLAMA_3_70B_INSTRUCT  // "meta-llama-3-70b-instruct"
GitHubModelsChatModelName.META_LLAMA_3_8B_INSTRUCT   // "meta-llama-3-8b-instruct"

Choose By Size:

• 405B: Maximum capability, complex reasoning
• 70B: Balanced capability and speed
• 8B: Fast, efficient, simple tasks

Best For: Open-source deployment, customization, various task complexities

Mistral Models

GitHubModelsChatModelName.MISTRAL_NEMO         // "Mistral-nemo"
GitHubModelsChatModelName.MISTRAL_LARGE        // "Mistral-large"
GitHubModelsChatModelName.MISTRAL_LARGE_2407   // "Mistral-large-2407"
GitHubModelsChatModelName.MISTRAL_SMALL        // "Mistral-small"

Best For: European languages, multilingual tasks

Choose:

• Large (2407): Latest large model
• Large: Standard large model
• Nemo: Efficient, balanced
• Small: Fast, simple tasks

Embedding Models

OpenAI Text Embedding Models

Text-Embedding-3-Small

GitHubModelsEmbeddingModelName.TEXT_EMBEDDING_3_SMALL  // "text-embedding-3-small"

Dimensions: 1536 (default), customizable Custom Dimensions: Yes (down to any size)

Best For:

• General-purpose semantic search
• Most applications
• Good balance of quality and performance
• Cost-effective

Performance: High quality, fast

Text-Embedding-3-Large

GitHubModelsEmbeddingModelName.TEXT_EMBEDDING_3_LARGE  // "text-embedding-3-large"

Dimensions: 3072 (default), customizable Custom Dimensions: Yes (down to any size)

Best For:

• Maximum embedding quality
• High-precision semantic search
• Complex similarity tasks
• Research applications

Performance: Highest quality, moderate speed

Dimension Tradeoff:

// Default: Maximum quality
.modelName(TEXT_EMBEDDING_3_LARGE)  // 3072 dimensions

// Reduced: Better performance, slight quality reduction
.modelName(TEXT_EMBEDDING_3_LARGE).dimensions(512)  // 512 dimensions

Cohere Embedding Models

Cohere-Embed-V3-English

GitHubModelsEmbeddingModelName.COHERE_EMBED_V3_ENGLISH  // "cohere-embed-v3-english"

Dimensions: 1024 (fixed) Custom Dimensions: No

Best For:

• English-only applications
• Optimized English understanding
• Fixed-dimension requirements

Cohere-Embed-V3-Multilingual

GitHubModelsEmbeddingModelName.COHERE_EMBED_V3_MULTILINGUAL  // "cohere-embed-v3-multilingual"

Dimensions: 1024 (fixed) Custom Dimensions: No Languages: 100+ languages

Best For:

• Multilingual applications
• Cross-language semantic search
• International applications

Model Selection Guide

By Task Complexity

Simple Tasks

• GPT-4o Mini
• Phi-3.5 Mini Instruct
• Meta Llama 3.1 8B
• Mistral Small

Moderate Tasks

• GPT-4o
• Meta Llama 3.1 70B
• Cohere Command R
• Mistral Large

Complex Tasks

• GPT-4o (latest features)
• Meta Llama 3.1 405B
• Mistral Large 2407

By Speed Requirements

Fast Response

• GPT-4o Mini
• Phi-3.5 Mini
• Meta Llama 3.1 8B

Balanced

• GPT-4o
• Meta Llama 3.1 70B
• Cohere Command R

By Context Length

Short Context (< 4K)

• Phi-3 Mini/Small/Medium 4K variants

Standard Context (8K-32K)

• Phi-3 8K variants
• Most models (default)

Long Context (128K+)

• GPT-4o (128K)
• GPT-4o Mini (128K)
• Phi-3 128K variants

By Special Capabilities

Vision Tasks

• GPT-4o (multimodal)
• Phi-3.5 Vision Instruct

Tool Calling

• GPT-4o
• GPT-4o Mini
• Most chat models

JSON Schema

• GPT-4o
• GPT-4o Mini
• Models with structured output support

Embedding Model Selection

General Use

TEXT_EMBEDDING_3_SMALL  // Best balance

Maximum Quality

TEXT_EMBEDDING_3_LARGE  // Highest quality

Multilingual

COHERE_EMBED_V3_MULTILINGUAL  // 100+ languages

Storage Optimization

TEXT_EMBEDDING_3_SMALL.dimensions(512)  // Reduced dimensions
COHERE_EMBED_V3_ENGLISH  // Fixed 1024 dimensions

Cost Considerations

Generally:

• Smaller models = Lower cost
• Shorter context = Lower cost
• Fewer tokens = Lower cost

Balance:

• GPT-4o Mini for volume
• GPT-4o for quality
• Llama models for cost-effective open source

Embeddings:

• TEXT_EMBEDDING_3_SMALL for balance
• Custom dimensions to reduce storage costs

Example Model Configurations

High-Quality Content Generation

GitHubModelsChatModel model = GitHubModelsChatModel.builder()
    .gitHubToken(token)
    .modelName(GitHubModelsChatModelName.GPT_4_O)
    .temperature(0.8)
    .maxTokens(2000)
    .build();

Fast Customer Support

GitHubModelsChatModel model = GitHubModelsChatModel.builder()
    .gitHubToken(token)
    .modelName(GitHubModelsChatModelName.GPT_4_O_MINI)
    .temperature(0.3)
    .maxTokens(500)
    .timeout(Duration.ofSeconds(30))
    .build();

Complex Reasoning

GitHubModelsChatModel model = GitHubModelsChatModel.builder()
    .gitHubToken(token)
    .modelName(GitHubModelsChatModelName.META_LLAMA_3_1_405B_INSTRUCT)
    .temperature(0.2)
    .maxTokens(3000)
    .build();

Multilingual Search

GitHubModelsEmbeddingModel model = GitHubModelsEmbeddingModel.builder()
    .gitHubToken(token)
    .modelName(GitHubModelsEmbeddingModelName.COHERE_EMBED_V3_MULTILINGUAL)
    .build();

Vision Tasks

GitHubModelsChatModel model = GitHubModelsChatModel.builder()
    .gitHubToken(token)
    .modelName(GitHubModelsChatModelName.PHI_3_5_VISION_INSTRUCT)
    .build();

Model Comparison Table

Chat Models

Embedding Models

See Also

• Chat Model API
• Embedding Model API
• Model Names API
• Best Practices