IQuest-Coder-V1-40B-Instruct-GGUF

Description

This repository contains GGUF format model files for IQuestLab's IQuest-Coder-V1-40B-Instruct.

IQuest-Coder-V1 is a state-of-the-art coding model built on a "code-flow" training paradigm. It captures the dynamic evolution of software logic, delivering exceptional performance on benchmarks like SWE-Bench Verified (81.4%) and BigCodeBench. This model natively supports a 128K context window.

Evaluation Results

Files & Quantization

To see the available files, please verify the Files and versions tab.

How to Run (llama.cpp)

Recommended Parameters: For coding tasks, a lower temperature is generally recommended for precision.

• Temperature: 0.2 (for code generation) or 0.7 (for creative explanations)
• Context: -c (Adjust based on your VRAM/RAM. The model supports up to 128k).

CLI Example

./llama-cli -m IQuest-Coder-V1-40B-Instruct.Q4_K_M.gguf \
  -c 8192 \
  --temp 0.2 \
  -p "User: Write a Python function to calculate the Fibonacci sequence using dynamic programming.\nAssistant:" \
  -cnv

Server Example

./llama-server -m IQuest-Coder-V1-40B-Instruct.Q4_K_M.gguf \
  --port 8080 \
  --host 0.0.0.0 \
  -c 16384 \
  -ngl 99