# Code generation

DiffuCoder-7B-cpGRPO-8bit is a code generation model converted to MLX format, based on apple/DiffuCoder-7B-cpGRPO, and is specifically designed to provide developers with an efficient code generation tool.

DiffuCoder-7B-cpGRPO-4bit is a 4-bit quantized version converted from the Apple DiffuCoder-7B-cpGRPO model, optimized for the MLX framework.

OpenThinker3-7B-GGUF is a quantized version of open-thoughts/OpenThinker3-7B, optimized for efficient inference. It is fine-tuned based on Qwen/Qwen2.5-7B-Instruct and performs excellently on mathematical, code, and scientific problems.

A code processing model fine-tuned based on Salesforce/codet5-base, focusing on Python code-related tasks

Seed-Coder-8B-Instruct is a powerful open-source code model with features such as model-centricity, transparency, and high performance, and it performs excellently in various coding tasks.

Spec-T1-RL-7B is a high-precision large language model focused on mathematical reasoning, algorithm problem-solving, and code generation, and it performs excellently in technical benchmark tests.

This is a 4-bit quantized version in MLX format converted from the Microsoft Phi-4-mini-reasoning model, suitable for text generation tasks.

Phi-4-reasoning is an advanced reasoning model fine-tuned based on Phi-4. Through supervised fine-tuning and reinforcement learning, it demonstrates excellent reasoning abilities in fields such as mathematics, science, and coding.

OlympicCoder-32B is a code generation model based on Qwen2.5-Coder-32B-Instruct, employing IQ-DynamicGate ultra-low-bit quantization technology for efficient inference in memory-constrained environments.

Theta - 35 is an advanced inference model in the Theta series launched by SVECTOR, focusing on complex thinking and reasoning, and excels in difficult problems that require in - depth logical analysis and multi - step reasoning.

DeepSeek-R1 is the first-generation inference model, which performs excellently in mathematics, code, and reasoning tasks, and its performance is comparable to that of OpenAI-o1.

IBM Granite 8B code instruction model, supporting a context length of 128k, focusing on code generation and instruction understanding tasks.

Qwen2.5-Coder-1.5B is a code generation model with 1.5B parameters, supporting multiple programming languages and suitable for code completion and generation tasks.

Based on the Qwen2.5-Coder-3B-Instruct model, quantization processing is performed, providing an efficient and convenient solution for code generation and dialogue interaction.

A large language model developed by the National Institute of Informatics in Japan, supporting multiple languages such as Japanese and English, and suitable for natural language processing tasks.

Italian language-optimized large language model based on Meta-Llama-3.1-8B-Instruct

A fine-tuned version of the Llama3.1 8B large language model optimized for Turkish, trained on high-quality Turkish instruction datasets

Mistral-Nemo-Base-2407 is a 12-billion-parameter generative text pre-training model jointly trained by Mistral AI and NVIDIA, outperforming models of similar or smaller scale.

Meta Llama 3 series large language model, featuring an 8 billion parameter instruction fine-tuned text generation model optimized for dialogue scenarios.

A Mixture of Experts (MoE) large language model developed by Databricks, with 132 billion total parameters and 36 billion active parameters, supporting a 32K context window

CroissantLLM is a 1.3B-parameter language model trained on 3T English-French bilingual tokens, designed for research and industrial applications, capable of running smoothly on consumer-grade hardware.

CodeLlama 70B Instruct is a large-scale code generation model based on the Llama 2 architecture, specifically optimized for code understanding and generation tasks.

Code Llama is a 70B-parameter Python-specialized code generation model developed by Meta, optimized based on the Llama-2 architecture with support for 16k context length

Piccolo-math-2x7b is a large language model specializing in mathematical and logical reasoning, named in honor of the author's pet dog Klaus. The model demonstrates outstanding performance across multiple benchmarks, particularly in mathematical and code generation tasks.

A 32k-context fine-tuned version based on TinyLlama-1.1B, achieving long-context processing capability by increasing rope theta

ARIA 70B V2 is a large-scale language model based on the Llama 2 architecture, supporting French and English, with a focus on text generation tasks.

WizardCoder Python 13B V1.0 is a large language model developed by WizardLM, focusing on Python code generation tasks. It is based on the llama2 architecture and performs excellently in the HumanEval benchmark test.

replit-code-v1-3b is a 2.7B parameter causal language model focused on code completion, developed by Replit, Inc.

Instruction-following model fine-tuned on LLaMA-7B, specifically optimized for code instruction scenarios

Pre-trained API recommendation generation model based on T5 Small architecture, specifically designed for Java programming tasks