Seed - Coder-8B-Reasoning-bf16 Open - Source Code Model: Enhanced Reasoning, Supports 64K Context Code Writing

Seed Coder 8B Reasoning Bf16

Developed by ByteDance-Seed

Seed-Coder is an 8B-scale open-source code model family, including base, instruction, and reasoning versions. The reasoning version enhances reasoning capabilities through reinforcement learning training and supports 64K context length.

Large Language Model

Transformers

Open Source License:MIT #64K long-context reasoning #Code generation optimization #Reinforcement learning training

Downloads 4,382

Release Time : 5/13/2025

Model Overview

Seed-Coder-8B-Reasoning-bf16 is a parameter-efficient 8B-scale open-source code model that enhances reasoning capabilities through reinforcement learning training and supports 64K context length.

Model Features

Model-centric data processing

Primarily utilizes LLMs rather than manual rules for code data filtering, minimizing human intervention in pre-training data construction.

Transparency and openness

Publicly shares detailed insights into the model-centric data pipeline, including methods for organizing GitHub data, commit data, and code-related web data.

High performance

Achieves state-of-the-art performance among open-source models of the same scale across various coding tasks.

Long-context support

Supports 64K context length, making it suitable for handling long code and complex reasoning tasks.

Model Capabilities

Code generation

Code completion

Code understanding

Complex reasoning

Use Cases

Software development

Code auto-completion

Helps developers write code quickly and improves development efficiency.

Code review

Analyzes code quality and identifies potential issues.

Education

Programming learning assistance

Helps students understand and learn programming concepts.

🚀 Seed-Coder-8B-Reasoning-bf16

Seed-Coder is a powerful, transparent, and parameter-efficient family of open - source code models at the 8B scale, featuring base, instruct, and reasoning variants. This is the bf16 version of the Seed - Coder-8B-Reasoning model, which shows excellent performance in various coding tasks.

✨ Features

We are thrilled to introduce Seed-Coder, a powerful, transparent, and parameter-efficient family of open-source code models at the 8B scale, featuring base, instruct, and reasoning variants. Seed-Coder contributes to promote the evolution of open code models through the following highlights.

Model-centric: Seed-Coder predominantly leverages LLMs instead of hand-crafted rules for code data filtering, minimizing manual effort in pretraining data construction.
Transparent: We openly share detailed insights into our model-centric data pipeline, including methods for curating GitHub data, commits data, and code-related web data.
Powerful: Seed-Coder achieves state-of-the-art performance among open-source models of comparable size across a diverse range of coding tasks.

This is the bf16 version of the Seed-Coder-8B-Reasoning model, which has the following features:

Property	Details
Model Type	Causal language models
Training Stage	Pretraining & Post-training
Data Source	Public datasets
Context Length	65,536

📦 Installation

You will need to install the latest versions of transformers and accelerate:

pip install -U transformers accelerate

🚀 Quick Start

Here is a simple example demonstrating how to load the model and perform code generation using the Hugging Face pipeline API:

Basic Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "ByteDance-Seed/Seed-Coder-8B-Reasoning-bf16"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True)

messages = [
    {"role": "user", "content": "Write a quick sort algorithm."},
]

input_ids = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    return_tensors="pt",
    add_generation_prompt=True,  
).to(model.device)

outputs = model.generate(input_ids, max_new_tokens=16384)
response = tokenizer.decode(outputs[0][input_ids.shape[-1]:], skip_special_tokens=True)
print(response)

📚 Documentation

Model Downloads

Model Name	Length	Download	Notes
Seed-Coder-8B-Base	32K	🤗 Model	Pretrained on our model-centric code data.
Seed-Coder-8B-Instruct	32K	🤗 Model	Instruction-tuned for alignment with user intent.
Seed-Coder-8B-Reasoning	64K	🤗 Model	RL trained to boost reasoning capabilities.
👉 Seed-Coder-8B-Reasoning-bf16	64K	🤗 Model	RL trained to boost reasoning capabilities.

Evaluation

Seed-Coder-8B-Reasoning strikes impressive performance on competitive programming, demonstrating that smaller LLMs can also be competent on complex reasoning tasks. Our model surpasses QwQ-32B and DeepSeek-R1 on IOI'2024, and achieves an ELO rating comparable to o1-mini on Codeforces contests.

For detailed benchmark performance, please refer to our 📑 Technical Report.

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご