Bean bag large model 1.8 ByteDance Seedance

Average daily call volume of 50 trillion tokens, ranking first in the world in BrowserComp benchmark test

50 trillion Daily average Tokens
#1 BrowserComp
1.8 Latest version
15s Video generation

🫘 What is a bean bag model?

Big bean bag model is a series of large language models launched by ByteDance’s Volcano Engine. Released in December 2025 Bean bag 1.8 surpassed Claude 3.5 and GPT-4o in the BrowserComp benchmark test to rank first in the world. At the same time, the average daily tokens calls exceeded 50 trillion, becoming one of the busiest large models in the world.

🏆 No. 1 in the world: Doubao 1.8 ranks first in the world with a success rate of 54.1% in the BrowserComp Agent benchmark test, surpassing Claude 3.5 Sonnet (51.6%) and GPT-4o (49.8%), demonstrating powerful web page automation and Agent capabilities.

🌐 Browser Agent

The world’s strongest web automation agent capability, BrowserComp #1

🎬 Seedance Video

Seedance 1.5 Pro supports 15-second high-quality video generation

📊 Super large scale

Daily average of 50 trillion tokens, serving products with hundreds of millions of users such as Douyin

🔧 Function call

Enhanced Function Calling capabilities to build complex applications

🚀 Model family

Bean Bag 1.8 Pro Ultimate version

The most powerful version, BrowserComp ranks first in the world, suitable for complex tasks and Agent applications.

54.1%
BrowserComp
128K
Context
Super strong
Agent capabilities

Seedance 1.5 Pro Video generation

Bytedance’s latest video generation model supports 15-second high-quality videos and is industry-leading in dynamic consistency.

15 seconds
Maximum duration
1080P
Resolution
24fps
Frame rate

Bean Bag 1.8 Lite Lightweight version

Cost-effective version, suitable for daily conversations and simple tasks, cost optimized.

fast
Response speed
32K
Context
low price
Cost

💻 API call example

Call the beanbag large model through the Volcano Engine API:

import requests

url = "https://ark.cn-beijing.volces.com/api/v3/chat/completions"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

data = {
    "model": "doubao-1-8-pro-32k",
    "messages": [
        {"role": "user", "content": "请介绍一下你自己"}
    ],
    "temperature": 0.7
}

response = requests.post(url, headers=headers, json=data)
print(response.json()["choices"][0]["message"]["content"])

💰 Price comparison

Model Input (¥/million tokens) Output (¥/million tokens) Features
Bean Bag 1.8 Pro ¥0.8 ¥2 Strongest performance
Bean Bag 1.8 Lite ¥0.3 ¥0.6 High cost performance
GPT-4o ¥18 ¥54 OpenAI Flagship
Claude 3.5 Sonnet ¥21 ¥105 Anthropic
💡 King of cost performance: The price of Doubao 1.8 Pro is only about 4% of GPT-4o, but its performance exceeds GPT-4o in the BrowserComp test, making it one of the most cost-effective large models currently.

🎯 Typical application scenarios

🌐 Web page automation

BrowserComp #1, Build a powerful web crawling and automation agent

🎬 Video creation

Seedance 1.5 Pro generates high-quality short videos to serve Douyin creators

📱 Intelligent customer service

Ultra-low cost deployment of enterprise-level intelligent customer service system

🔍 Content understanding

Multi-modal understanding ability, processing mixed content of graphics and text