Model Catalog

Browse and compare AI models available for your Thox.ai device.

Available Models

Thox.ai comes pre-loaded with optimized AI models. You can download additional models based on your needs.

Thox Coder

thox-coder

Optimized for code completion, generation, and refactoring tasks.

Size

7B parameters

Memory

6GB

Speed

45 tok/s

Best for

Code completionBug fixingCode explanationUnit test generation

Thox Coder Large

thox-coder-large

Enhanced coding model with broader knowledge and better reasoning.

Size

13B parameters

Memory

10GB

Speed

28 tok/s

Best for

Complex algorithmsArchitecture designCode reviewMulti-file refactoring

Thox Chat

thox-chat

General-purpose conversational AI for everyday tasks.

Size

7B parameters

Memory

6GB

Speed

50 tok/s

Best for

ConversationsWriting assistanceBrainstormingQ&A

Thox Chat Large

thox-chat-large

Advanced conversational model with improved reasoning capabilities.

Size

13B parameters

Memory

10GB

Speed

32 tok/s

Best for

Complex reasoningAnalysisCreative writingDetailed explanations

Thox Embed

thox-embed

Fast embedding model for semantic search and RAG applications.

Size

335M parameters

Memory

1GB

Speed

2000 tok/s

Best for

Semantic searchDocument similarityRAG pipelinesClustering

Thox Instruct

thox-instruct

Instruction-following model optimized for task completion.

Size

7B parameters

Memory

6GB

Speed

48 tok/s

Best for

Task completionData extractionSummarizationFormat conversion

Installing Models

Via CLI

# List available models

thox models list --remote

# Install a model

thox models pull thox-coder-large

# Check installed models

thox models list

Via Web Interface

  1. 1Navigate to http://thox.local:8080/admin
  2. 2Go to the "Models" section
  3. 3Click "Download" next to your desired model

Quantization Options

Each model is available in different quantization levels, trading off quality for speed and memory usage.

QuantizationQualitySpeedMemory
fp16HighestSlowest~14GB (7B)
q8_0Very HighFast~7GB (7B)
q4_k_mGoodVery Fast~4GB (7B)
q4_0AcceptableFastest~3.5GB (7B)

Recommendation: For most users, q4_k_m offers the best balance of quality and performance. Use q8_0 if you need higher quality and have memory to spare.