tepirale/merge-cord-v1

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 9, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

tepirale/merge-cord-v1 is a 1.5 billion parameter language model created by tepirale, merged using the Consensus-Orthogonal Residual Decomposition (CORD) method. Based on Qwen2.5-1.5B, it integrates capabilities from both Qwen2.5-Coder-1.5B-Instruct and Qwen2.5-1.5B-Instruct. This model is designed to combine general instruction-following with specialized coding abilities, offering a balanced performance for various generative AI tasks within its 32768 token context window.

Loading preview...

Model Overview

tepirale/merge-cord-v1 is a 1.5 billion parameter language model developed by tepirale, leveraging the Consensus-Orthogonal Residual Decomposition (CORD) merge method. This model is built upon the Qwen/Qwen2.5-1.5B base and integrates the strengths of two distinct instruction-tuned models: Qwen/Qwen2.5-Coder-1.5B-Instruct and Qwen/Qwen2.5-1.5B-Instruct. The merge aims to combine general-purpose instruction following with enhanced coding capabilities, making it a versatile option for its size class.

Merge Details

The model was created using mergekit with a specific YAML configuration that defines the CORD method. This process involved combining layers from the base Qwen2.5-1.5B model with those from its instruction-tuned and coder-specific variants. The configuration specifies a consensus_threshold and lambda_consensus to guide the merging process, aiming for a balanced integration of features from the constituent models.

Key Characteristics

  • Parameter Count: 1.5 billion parameters, offering a compact yet capable model.
  • Context Length: Supports a substantial context window of 32768 tokens, suitable for processing longer inputs and generating more extensive outputs.
  • Hybrid Capabilities: Designed to handle both general instruction-following tasks and coding-related queries, benefiting from the specialized training of its merged components.

Intended Use Cases

This model is suitable for applications requiring a balance of general language understanding and code generation/comprehension, particularly where resource efficiency (due to its 1.5B size) and a large context window are important. It can be used for tasks such as:

  • General text generation and summarization.
  • Instruction-following and conversational AI.
  • Basic code generation, explanation, and debugging assistance.

Finetunes
1 models