🔥 FlintAPI — One API Key for 25+ Chinese LLMs

Access DeepSeek, Qwen, Kimi, GLM, MiniMax, and 20+ more models
through a single OpenAI-compatible API. $2 free credit — no card required.

---

Why FlintAPI?

	FlintAPI	Direct Provider	OpenRouter
25+ Chinese LLMs	✅ One API key	❌ Per-provider account	✅
Self-hosted PPU Qwen	✅ Lower cost	❌ Cloud pricing	❌
OpenAI-compatible	✅ Drop-in replacement	⚠️ Varies	✅
Free credit	✅ $2, no card	❌	⚠️ Limited
No middleman markup	✅ Self-hosted	✅	❌ Markup
Referral bonus	✅ $1 each	❌	❌

---

Quick Start (60 seconds)

1. Register

Go to flintapi.ai/register — you get $2 free credit instantly, no credit card required.

2. Get your API Key

Find it in your Dashboard → Settings → Create Key. Copy it — it's shown only once!

3. Make your first call

cURL:

curl https://flintapi.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "flint-smart",
    "messages": [{"role": "user", "content": "Explain quantum computing in one sentence."}]
  }'

Python (with pip install openai):

from openai import OpenAI

client = OpenAI(
    base_url="https://flintapi.ai/v1",
    api_key="YOUR_API_KEY"
)

response = client.chat.completions.create(
    model="flint-smart",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

Node.js (npm install openai):

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://flintapi.ai/v1',
  apiKey: 'YOUR_API_KEY',
});

const completion = await client.chat.completions.create({
  model: 'qwen3.7-max',
  messages: [{ role: 'user', content: 'Say hello in Chinese' }],
});
console.log(completion.choices[0].message.content);

Go (go get github.com/sashabaranov/go-openai):

import "github.com/sashabaranov/go-openai"

client := openai.NewClient("YOUR_API_KEY")
client.BaseURL = "https://flintapi.ai/v1"

resp, _ := client.CreateChatCompletion(ctx, openai.ChatCompletionRequest{
    Model: "glm-5.1",
    Messages: []openai.ChatCompletionMessage{
        {Role: "user", Content: "Hello!"},
    },
})
fmt.Println(resp.Choices[0].Message.Content)

---

🧠 flint-smart Router

Don't know which model to pick? Use "model": "flint-smart" and our router auto-selects the best Chinese LLM for your prompt:

Prompt Type	Routed To	Why
Code / programming	deepseek-v4-pro	Best coding LLM
Reasoning / analysis	deepseek-v4-pro	Strong logic
Chinese / bilingual	deepseek-v4-pro	Native Chinese
General / default	deepseek-v4-flash	Fast & cheap

---

📊 Available Models (18+)

Model	Provider	Context	Best For
deepseek-v4-pro	DeepSeek	128K	Reasoning & code
deepseek-v4-flash	DeepSeek	128K	Fast, cost-effective
qwen3.7-max	Qwen	128K	Flagship all-rounder
qwen3.7-plus	Qwen	128K	Balanced Chinese-English
qwen3.5-plus	Qwen	128K	Solid general-purpose
qwen3.5-flash	Qwen	128K	Lowest latency
kimi-k2.6	Kimi	256K	Long context
kimi-k2.5	Kimi	128K	Long context
glm-5.1	GLM	32K	Bilingual expert
MiniMax-M2.7	MiniMax	512K	Creative writing

Full list & pricing: flintapi.ai/pricing

---

🔌 Python SDK

For an even simpler experience:

pip install flintapi

from flintapi import Flint

flint = Flint(api_key="YOUR_API_KEY")

# Simple chat
reply = flint.chat("Explain quantum computing in one sentence.")
print(reply)

# With model selection
reply = flint.chat("Write a Python quicksort", model="flint-smart")
print(reply)

# Streaming
for chunk in flint.chat_stream("Tell me a story"):
    print(chunk, end="")

---

📖 Docs & Resources

• API Docs — Full API reference with code examples
• Playground — Try models in your browser
• Pricing — Per-token pricing for all models
• Compare — FlintAPI vs alternatives
• Status — Real-time model health
• Dashboard — Usage, billing, API keys

---

🌟 Features

• One API key for 25+ Chinese LLMs
• OpenAI-compatible — change base_url and keep existing code
• flint-smart router — auto-select best model per request
• $2 free credit — instant, no card required
• Real-time billing — per-token pricing, usage dashboard
• 4 language SDKs — cURL, Python, Node.js, Go
• PPU self-hosted Qwen — competitive pricing on custom silicon
• Referral program — both you and your friend get $1

---

🤝 Support

• Email: support@flintapi.ai
• GitHub Issues: github.com/moozechen/flintapi/issues
• Website: flintapi.ai

---

🇨🇳 Chinese AI Landscape (2026)

A community resource tracking the Chinese LLM ecosystem. Not promotional — just what's out there, what each model does best, and how to try them.

The Major Players

Model	Organization	Open Weights	Best For	How to Try
DeepSeek-V4-Pro	DeepSeek	✅	Reasoning, math, code	api-docs.deepseek.com
Qwen3.7-Max	Alibaba	✅	All-rounder, bilingual	qwen.ai
Kimi-K2.6	Moonshot AI	❌ API only	Long context (256K)	kimi.moonshot.cn
GLM-5.1	Zhipu AI	✅	Bilingual enterprise	zhipuai.cn
MiniMax-M2.7	MiniMax	❌ API only	Creative writing, 512K ctx	minimaxi.com
Doubao-1.5-pro	ByteDance	❌ API only	Chinese content, fast	volcengine.com
Baichuan-M1	Baichuan	✅	Healthcare, finance	baichuan-ai.com
Yi-Lightning	01.AI	✅	Efficient, low-latency	01.ai

Why They Matter

• DeepSeek-V4 scored #1 on Aider's polyglot coding benchmark, beating GPT-5 and Claude
• Qwen3.6-35B-A3B (MoE) delivers near-70B performance at 35B params — 1274 points on HN
• Qwen3.6-27B runs on a single RTX 3090 (Q4 quant) at 50-70 tok/s — HN front page
• Kimi-K2.6 handles 256K context windows, competitive with Gemini for long-document tasks
• MiniMax-M2.7 has 512K context — among the longest available worldwide

Community Benchmarks

Benchmark	Top Chinese Model	Score vs GPT-5
Aider Polyglot (coding)	DeepSeek-V4-Pro	+5.2%
LiveCodeBench	Qwen3.7-Max	-2.1%
MMLU-Pro	DeepSeek-V4-Pro	-1.8%
C-Eval (Chinese NLP)	Qwen3.7-Max	+8.3%

Benchmarks from official model cards and public leaderboards. Last updated: June 2026.

Getting Started

1. Via API: Most providers offer OpenAI-compatible endpoints — sign up, get a key, swap base_url
2. Run locally: Qwen3.6-27B fits on a 3090; DeepSeek-V4 needs ~140GB VRAM (run on cloud or multiple GPUs)
3. Registration tip: Some platforms require Chinese phone numbers — use Alibaba Cloud's international portal for Qwen without Chinese ID

---

Community-maintained resource. Spotted outdated info? Open an issue.