Unixcoder Base
Microsoft's Unixcoder Base is a unified cross-modal pre-trained model that leverages multimodal data to pretrain code representation, particularly exceling at tasks such as code search, code completion, and code summarization. Its architecture is based on the RoBERTa model, indicating a robust foundation for natural language processing.
Input
Output
Context
1K
Max Output
0K
Parameters
-
Technical Specifications
Model TypeEmbedding
Context Window512 tokens
Max Output Tokens128 tokens
ParametersNot available
Release DateMar 23, 2022
Training CutoffNot available
Licenseapache-2.0
Open SourceYes
Input Modalities
Text
Output Modalities
Embedding
Capabilities
Resources & Links
More Models by Microsoft
Table Transformer Structure Recognition V1.1 All28.8M · Open weightsPhi 4 Mini Instruct131K context · 3.8B · Open weightsDeberta Large Mnli1K context · Open weightsBeit Base Patch 16 224Open weightsVibevoice ASR HF131K context · 8.3B · Open weightsPhi 3.5 Vision Instruct131K context · 4.1B · Open weights
Related Tools
Data sourced from official provider APIs and documentation
Last updated: Aug 8, 2026
Your AI stack shouldn't stand still.
Every month new models become cheaper, faster, and more capable. Inferbase ensures your application automatically benefits without changing a single API call.