mcp-tool-use-quality-ranger-0.6b-GGUF

mcp-tool-use-quality-ranger-0.6b is a sequence classification model fine-tuned from Qwen3-0.6B-Base, designed to evaluate the quality of function calls within conversational AI systems using the Model Context Protocol (MCP) framework. Supporting a context length of 32,768 tokens, it classifies function calls as VALID_CALL, TOOL_ERROR, PARAM_NAME_ERROR, or PARAM_VALUE_ERROR by verifying tool selection, parameter names, and parameter values, and delivers robust, fast assessments for dialog-based tool usage, parameter errors, and value correctness. Optimized for lightweight deployment, mcp-tool-use-quality-ranger-0.6b achieves high benchmark accuracy, making it ideal for developers and researchers who require precise tool call evaluations in AI workflows.

Model Files

File Name	Quant Type	File Size
mcp-tool-use-quality-ranger-0.6b.BF16.gguf	BF16	1.2 GB
mcp-tool-use-quality-ranger-0.6b.F16.gguf	F16	1.2 GB
mcp-tool-use-quality-ranger-0.6b.F32.gguf	F32	2.39 GB
mcp-tool-use-quality-ranger-0.6b.Q2_K.gguf	Q2_K	296 MB
mcp-tool-use-quality-ranger-0.6b.Q3_K_L.gguf	Q3_K_L	368 MB
mcp-tool-use-quality-ranger-0.6b.Q3_K_M.gguf	Q3_K_M	347 MB
mcp-tool-use-quality-ranger-0.6b.Q3_K_S.gguf	Q3_K_S	323 MB
mcp-tool-use-quality-ranger-0.6b.Q4_0.gguf	Q4_0	382 MB
mcp-tool-use-quality-ranger-0.6b.Q4_1.gguf	Q4_1	409 MB
mcp-tool-use-quality-ranger-0.6b.Q4_K.gguf	Q4_K	397 MB
mcp-tool-use-quality-ranger-0.6b.Q4_K_M.gguf	Q4_K_M	397 MB
mcp-tool-use-quality-ranger-0.6b.Q4_K_S.gguf	Q4_K_S	383 MB
mcp-tool-use-quality-ranger-0.6b.Q5_0.gguf	Q5_0	437 MB
mcp-tool-use-quality-ranger-0.6b.Q5_1.gguf	Q5_1	464 MB
mcp-tool-use-quality-ranger-0.6b.Q5_K.gguf	Q5_K	444 MB
mcp-tool-use-quality-ranger-0.6b.Q5_K_M.gguf	Q5_K_M	444 MB
mcp-tool-use-quality-ranger-0.6b.Q5_K_S.gguf	Q5_K_S	437 MB
mcp-tool-use-quality-ranger-0.6b.Q6_K.gguf	Q6_K	495 MB
mcp-tool-use-quality-ranger-0.6b.Q8_0.gguf	Q8_0	639 MB

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):