8341TokenFactory
NEWGlobal flagship models are now live — 6 world-leading models, one API key for allSee models & pricing →
8341TokenFactory — All services operational

Compute-Power-Carbon Synergy AI Computing Operator

An AI LLM API aggregation platform by token8341

Model

Which model do you want to use? Tell me and I'll route it — no signup needed

Need a model we don't have yet? Say so — we'll add it fast

Free · No signup · 30 req/minAPI pricing →
免费领 API KeyView Pricing

8341TokenFactory is the commercialization platform of the Juyongguan Key Laboratory of Energy Security and Intelligent Technology (provincial and ministerial level) at North China Electric Power University, built by a technical team from Peking University and Tsinghua University. It aggregates leading models such as DeepSeek, GLM-5.3, Kimi K3, MiniMax H3 and Qwen3-Coder, along with foreign flagships including Claude, Gemini and GPT. Three in-house technologies hold costs down: compute-power co-scheduling, inference acceleration, and super-intelligent routing. An AI token and LLM API aggregation platform: lower cost, one API key for every model — making AI affordable, and making AI clean. The platform operates global service nodes and compute centers in Singapore, Seoul, Silicon Valley, Frankfurt, and in Beijing, Shanghai, Hangzhou, Guangzhou, Shenzhen, Ulanqab, Horinger and Zhongwei, China.

Supported ModelsDeepSeek V4.1 Flash 1017Kimi K3Qwen 3.8 MaxGLM-5.3MiniMax H3DeepSeek-V4-ProQwen3-CoderDoubao 2.1

Supported Models

Frontier models in one API — reasoning, chat, code, image and video generation.

International Models

C

Claude Opus 5.5

Anthropic flagship reasoning model for long-context and complex tasks

C

Claude Sonnet 5.5

Cost-effective flagship for coding and tool calling

G

GPT-6.1 Sol

OpenAI flagship with balanced general and reasoning ability

G

GPT-6 Astra

OpenAI high-throughput model for long-form and multi-turn chat

G

Gemini 4 Argon

Google next-gen multimodal model with strong vision-language understanding

G

Gemini 3.8 Flash

Google low-latency model for high-concurrency workloads

Chinese Models

D

DeepSeek V4.1 Flash 1017

DeepSeek V4.1 Flash — 1M context, thinking mode and image understanding, from ¥0.8/M input tokens

K

Kimi K3

Moonshot Kimi K3 / K2.7-Code — ultra-long context and strong coding ability

Q

Qwen-Max

Alibaba Qwen — exceptional Chinese language understanding and generation

G

GLM-5.3

Zhipu GLM-5.3 — significantly improved long-horizon task and coding performance, supports ultra-long context

M

MiniMax H3

MiniMax H3 multimodal video generation model, supporting text-to-video, image-to-video, and reference generation

D

DeepSeek-V4-Pro

Cost-effective open-source model with strong reasoning performance

D

Doubao

ByteDance Doubao — multimodal understanding, high concurrency with low latency

E

ERNIE 4.0

Baidu ERNIE 4.0 — knowledge-enhanced LLM, deeply optimized for Chinese scenarios

S

Spark 4.0

iFlytek Spark 4.0 — cognitive model with speech-text multimodal fusion

P

Pangu 5.0

Huawei Pangu 5.0 — China sovereign AI, full-stack capabilities across vision, language, and reasoning

GLOBAL NODES

Global Service Nodes & Compute Centers

We operate global service nodes in Singapore, Seoul, Silicon Valley, Frankfurt, and in Beijing, Shanghai, Hangzhou, Guangzhou and Shenzhen, China. Our compute centers are located in Ulanqab and Horinger in Inner Mongolia, and Zhongwei in Ningxia.

12
Nodes
5
Regions
12
Cities
🇨🇳
China
Beijing · Shanghai · Hangzhou · Guangzhou · Shenzhen
🌏
Asia-Pacific
Singapore · Seoul
🌎
North America
Silicon Valley
🌍
Europe
Frankfurt
Compute Centers
Ulanqab · Horinger (Inner Mongolia) · Zhongwei (Ningxia)
Service Nodes
Singapore · Seoul · Silicon Valley · Frankfurt · Beijing · Shanghai · Hangzhou · Guangzhou · Shenzhen

Global service nodes and compute centers in Singapore, Seoul, Silicon Valley, Frankfurt, and Beijing, Shanghai, Hangzhou, Guangzhou and Shenzhen, China.

AI Video Generation

One key to call multiple video models — text-to-video, image-to-video, first/last-frame, and reference-image generation. Billed per second, OpenAI-compatible.

Multiple models: text-to-video / image-to-video / first-last frame / reference image
Billed per second, from 4 seconds, transparent pricing
OpenAI-compatible — switch video models with one line of code
M

MiniMax H3

MiniMax H3 — multimodal video generation: text-to-video, image-to-video, and reference generation

W

Wan 2.2

Wan 2.2 — text-to-video, cost-effective batch video production

View video models

AI Short-Drama Agent

From one sentence to a finished video: script, storyboard, shots, and voiceover — fully automated. Ideal for short dramas, ads, and social media content.

Generate script and storyboard from one sentence
Character consistency + lip sync
Automatic voiceover and music, ready-to-publish videos
Learn more →

Why Choose 8341TokenFactory

Compute-Power-Carbon Synergy, great pricing - the best AI API aggregation platform to make LLM API access accessible and affordable

Competitive Pricing

Direct partnerships with model providers, volume discounts, pay only for what you use.

Fast & Reliable

Beijing, Ulanqab, Weihai (Shandong), Shaoguan (Guangdong), Guizhou, Horinger, Zhongwei (Ningxia), and other computing power centers. Self-developed computing power scheduling algorithms enable cross-regional intelligent routing, multi-node redundant deployment, and SLA 99.9%.

One-Click Integration

Fully compatible with OpenAI API format, zero-code migration. Multi-language SDKs available.

Video & Short Drama

One key covers both LLM APIs and AI video generation; the short-drama Agent goes from script to finished video automatically — no need to switch platforms.

landing.advantage5Title

landing.advantage5Desc

One Line of Code, Instant Access

OpenAI SDK compatible, zero-code migration. Choose your language, copy the code, replace YOUR_API_KEY.

// Replace YOUR_API_KEY with your API Key
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.token8341.com/v1"
)

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True
)

for chunk in response:
    print(chunk.choices[0].delta.content or "", end="")
📦 New users get ¥0.5 free trial credit; get ¥5 extra on your first top-up of ¥5 or more — no credit card required

Quick Start

Four steps to start using AI models

01

Create Account

Fill in basic info, register in 30 seconds, start your AI journey.

02

Add Balance

Multiple payment methods, pay-as-you-go, balance available instantly.

03

Create API Key

Generate API keys in the dashboard with flexible permission controls.

04

Call Models

Use standard OpenAI SDK or HTTP requests to call various AI models.

Serving numerous developers and businesses
Reliable API access, always available
99.9%
Service Uptime

Frequently Asked Questions

Common questions about accessing, pricing and running LLM APIs on 8341TokenFactory

What is 8341TokenFactory?

8341TokenFactory is the commercialization platform of the Juyongguan Key Laboratory of Energy Security and Intelligent Technology (a provincial and ministerial-level key laboratory) at North China Electric Power University, with a technical team from Peking University and Tsinghua University. It unifies leading frontier models (DeepSeek, GLM-5.3, Kimi K3, MiniMax H3, Qwen3-Coder), foreign flagship models (Claude, Gemini, GPT) and Chinese LLMs (Pangu, Qwen, ERNIE, Doubao, Spark) behind a single OpenAI-compatible API, so one API key calls every model, billed by token usage. The cost advantage comes from three in-house technologies: a Compute-Power Co-scheduling Engine that schedules inference into the cheapest power windows, LoRe for inference acceleration and GPU memory reduction, and TokenRouter token-level routing (developed with a Tsinghua University team) that sends simple requests to small models and escalates only hard ones — cutting token consumption at the same quality.

Which large language models are supported?

The platform covers mainstream domestic and overseas models, including Pangu 5.0, DeepSeek-V4-Pro, Qwen-Max, ERNIE 4.0, Doubao, Spark 4.0, GLM-5.3, Kimi K3, MiniMax H3 and Qwen3-Coder, covering chat, reasoning, code generation, video and image generation. The catalogue is updated continuously; the current list is on the pricing page.

How is it priced? Is there a monthly fee?

There is no monthly fee and no minimum spend. You pay only for actual token usage. Because the platform buys capacity from model vendors in bulk, unit prices are lower than buying directly from each vendor, and cached requests cost even less. Balance tops up instantly and usage is itemised.

Is it OpenAI-compatible? How much code must I change?

Fully compatible. Projects built on the OpenAI SDK need two changes only: point base_url at the 8341TokenFactory endpoint, and replace api_key with a key issued by 8341TokenFactory. No other code, SDK or parameter changes are required — a zero-code migration.

How long does it take to get started?

Around three minutes: register (about 30 seconds), top up or use the new-user free credit, create an API key in the console, then paste the sample code with your key to make the first call. cURL, Python and Node.js samples are provided in the console.

Is there a free tier for new users?

Yes. New accounts receive free trial credit on registration — no credit card required — so you can test the service before topping up.

Is my data secure? Is it retained or used for training?

Requests and responses are transmitted over HTTPS. The platform does not persist call content and does not use customer data for model training. Enterprises with stricter compliance requirements can contact support for a commercial agreement.

How reliable is the service? Is there an SLA?

Service availability is committed at 99.9%. Compute is distributed across multiple data centres in Beijing, Weihai, Shaoguan, Guizhou, Horinger, Hami and Zhongwei. A proprietary scheduling algorithm routes requests intelligently across regions and fails over automatically when a node is unhealthy.

Which payment methods are supported? Can I get an invoice?

Multiple online payment methods are supported and funds arrive instantly. Business customers can request a VAT invoice. Contact the support assistant "Wu Qiong" for the exact process.

Who do I contact if something goes wrong?

Contact the online support assistant "Wu Qiong", or raise a ticket in the console and the technical team will investigate. Common API error codes and troubleshooting steps are also documented on the docs site.

Start with 8341TokenFactory Today — AI API Gateway & LLM Platform

Sign up for free trial credits, no credit card required, start your AI development journey with the best multi-model API platform

Free Sign Up