majimenez/broken-model-fixed
The majimenez/broken-model-fixed is an 8 billion parameter Qwen3-based causal language model with a 32768 token context length. This model is a corrected version of yunmorning/broken-model, specifically addressing a missing `chat_template` in its tokenizer configuration. The fix enables proper function of `/chat/completions` API endpoints and supports advanced features like tool calling and reasoning blocks, making it suitable for chat-based applications requiring structured interactions.
Loading preview...
Overview
The majimenez/broken-model-fixed is a corrected version of the yunmorning/broken-model model, which previously failed to serve functional /chat/completions API requests. The core issue was a missing chat_template field in the tokenizer_config.json file, preventing inference servers from correctly formatting chat messages into the model's expected prompt structure.
This fixed model, based on the Qwen3-8B architecture, now includes the standard Qwen3-8B chat template. This critical addition resolves the API functionality issues and enables advanced features that rely on proper chat formatting.
Key Capabilities & Fixes
- Restored
/chat/completionsAPI functionality: The primary fix ensures the model can now process chat messages correctly. - Enabled Tool Calling: The added
chat_templateincludes formatting logic for<tool_call>and<tool_response>tags, allowing for tool/function calling capabilities. - Supported Reasoning Blocks: The template also incorporates
<think>tags, facilitating the use of reasoning blocks within responses. - Corrected Metadata: The
README.mdwas updated to accurately reflect the model's Qwen3-8B base, correcting an erroneousbase_modeldeclaration.
Good For
- Chat-based applications: Ideal for use cases requiring robust
/chat/completionsAPI support. - Applications needing structured interactions: Benefits from the enabled tool calling and reasoning block features.
- Developers seeking a functional Qwen3-8B variant: Provides a reliable and correctly configured Qwen3-8B model for deployment.