🚀 Gemini 4 Argon — Unlimited Free API

The first open-source, OpenAI-compatible proxy engine that unlocks Google's next-generation Gemini 4 Argon frontier model with customizable Thinking Levels (High, Medium, Low) — 100% Free & Unlimited.

---

🌟 Why Gemini 4 Argon?

Gemini 4 Argon is Google's newest, highly-advanced frontier intelligence model engineered specifically for complex tasks, deep logic deduction, and emergent reasoning.

Until now, Argon's thinking controls were locked behind Google's internal web interface. This repository reverse-engineers Google's proprietary JSPB extension protocol to expose full programmatic control over Gemini 4 Argon via standard OpenAI-compatible REST and streaming endpoints:

• 🧠 Argon High (argon-high / Level 2): Deep multi-pass thinking trace for hard logic riddles, competitive programming, and complex architecture.
• ⚖️ Argon Medium (argon-medium / Level 4): Balanced thinking depth for structured debugging and comprehensive analysis.
• ⚡ Argon Low (argon-low / Level 1): Rapid reasoning with quick self-verification.
• 🚀 Gemini 3.8 Flash (Default Flagship): Powered by gemini-3.8-flash as the ultra-fast default engine (is_default: true) for everyday zero-latency tasks.

---

🔥 Key Highlights

Feature	Description
🧠 Gemini 4 Argon Native	Programmatic access to Google's frontier Argon model with multi-level reasoning control (argon-high, argon-medium, argon-low).
⚡ Gemini 3.8 Flash Default	gemini-3.8-flash is set as the default model (is_default: true) for instant sub-second completions without configuration.
🔌 Drop-In OpenAI Standard	Seamlessly connects with Cursor, Cline, NextChat, OpenWebUI, LangChain, and the standard OpenAI Python / JS SDKs.
🛡️ Distributed Account Pooling	Connect 1 to 100+ Google accounts to load-balance requests with least-busy routing and automatic rate-limit cooldown management.
🍪 Zero-Touch Cookie Bridge	Companion Chrome extension automatically discovers the local server and syncs session credentials with zero tab manipulation.
🚀 Pure Go HTTP/3 QUIC	State-of-the-art TLS 1.3 fingerprint simulation (Chrome 152 JA4) with 0-RTT session resumption for ultra-low latency.
📊 Dual Intelligence Analytics	Real-time SQLite WAL storage tracking token usage, latency, and estimated cost-savings exported directly to JSON and Excel (.xlsx).

---

📊 Supported Models & Thinking Matrix

Model ID	Aliases	Thinking Depth	Tier	Role & Description
gemini-4-argon	argon, 4-argon, gemini-4	Low (1)	Free / Auth	🌟 Frontier Intelligence for complex problem solving
gemini-4-argon-thinking-high	argon-high, thinking-high	High (2)	Free / Auth	🧠 Deep Reasoning with maximum thinking steps
gemini-4-argon-thinking-medium	argon-medium, thinking-medium	Medium (4)	Free / Auth	⚖️ Balanced Reasoning for code reviews and analysis
gemini-4-argon-thinking-low	argon-low, thinking-low	Low (1)	Free / Auth	⚡ Fast Thinking with lightweight verification
gemini-3.8-flash	flash, 3.8-flash, default	None	Free	🚀 Default Model — Ultra-fast, balanced multimodal flagship
gemini-3.8-flash-thinking-high	flash-high	High (2)	Free	⚡ Flash with deep thinking enabled
gemini-3.1-pro	pro, 3.1-pro	None	Subscription	💻 Advanced multi-turn coding (Gemini Advanced accounts)
gemini-3.1-pro-thinking-high	pro-high	High (2)	Subscription	🧠 Pro with deep multi-pass reasoning

---

⚡ Quickstart

1. Launch the Server

Option A: Native Go (Recommended)

git clone https://github.com/kodelyx/free-gemini-api.git
cd free-gemini-api/free-gemini-api
go run main.go

Option B: Docker Compose

cd free-gemini-api/free-gemini-api
docker compose up -d --build

The server listens on http://127.0.0.1:8001 (API) and ws://127.0.0.1:9226 (Cookie Bridge).

---

2. Connect Your Account (Zero-Touch)

1. Open chrome://extensions/ in Chrome.
2. Toggle Developer mode on (top-right).
3. Click Load unpacked and select the gemini-extension/ folder.
4. Log into gemini.google.com.
5. The extension automatically detects your session and streams the worker credentials to the local API pool.

---

💻 Code Examples

1. Gemini 4 Argon Deep Thinking (cURL)

curl -X POST http://localhost:8001/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer any-key" \
  -d '{
    "model": "argon-high",
    "messages": [
      {
        "role": "user",
        "content": "A farmer has 17 sheep. All but 9 die. How many sheep are left alive? Show your step-by-step thinking."
      }
    ]
  }'

2. Python (Official OpenAI SDK)

from openai import OpenAI

# Point client to your local Free Gemini 4 Argon server
client = OpenAI(
    base_url="http://localhost:8001/v1",
    api_key="not-needed"
)

# 1. Ask Gemini 4 Argon with High Thinking
response = client.chat.completions.create(
    model="argon-high",
    messages=[
        {"role": "system", "content": "You are a world-class algorithm designer."},
        {"role": "user", "content": "Write a thread-safe LRU cache in Go with O(1) eviction."}
    ]
)
print("🧠 Gemini 4 Argon:", response.choices[0].message.content)

# 2. Use Default Fast Model (Gemini 3.8 Flash)
flash_resp = client.chat.completions.create(
    model="gemini-3.8-flash",
    messages=[{"role": "user", "content": "Explain quantum superposition in 1 sentence."}]
)
print("⚡ Gemini 3.8 Flash:", flash_resp.choices[0].message.content)

3. JavaScript / TypeScript (Node & Bun)

import OpenAI from "openai";

const openai = new OpenAI({
  baseURL: "http://localhost:8001/v1",
  apiKey: "free-gemini",
});

async function main() {
  const stream = await openai.chat.completions.create({
    model: "argon-high",
    messages: [{ role: "user", content: "Solve: If 5 machines make 5 widgets in 5 minutes, how long do 100 machines take to make 100 widgets?" }],
    stream: true,
  });

  for await (const chunk of stream) {
    process.stdout.write(chunk.choices[0]?.delta?.content || "");
  }
}

main();

---

🛠️ Third-Party Client Integrations

Client / Tool	Base URL	Model Name	API Key
Cursor / Cline	http://localhost:8001/v1	argon-high or gemini-3.8-flash	Any string
NextChat / ChatGPT-Next-Web	http://localhost:8001/v1	argon-high, flash-high	Any string
OpenWebUI / LibreChat	http://localhost:8001/v1	gemini-4-argon, default	Any string
LangChain / LlamaIndex	http://localhost:8001/v1	argon-high	Any string

---

🏛️ System Architecture

                 [ Clients / Cursor / OpenAI SDK / Agents ]
                                     │
                                     ▼
                ┌─────────────────────────────────────────┐
                │        Free Gemini API Server           │
                │   (Port 8001: HTTP/3 | Port 9226: WS)   │
                └────────────────────┬────────────────────┘
                                     │
                        ┌────────────┴────────────┐
                        ▼                         ▼
             ┌─────────────────────┐   ┌─────────────────────┐
             │  Model Registry     │   │   Output Arranger   │
             │  - gemini-4-argon   │   │  (Native Tool Calls │
             │  - argon-high       │   │   & Sanitization)   │
             │  - gemini-3.8-flash │   └─────────────────────┘
             └──────────┬──────────┘
                        │
                        ▼
             ┌─────────────────────┐
             │ Multi-Account Pool  │
             │ (Least-Busy Worker) │
             └──────────┬──────────┘
                        │
         ┌──────────────┴──────────────┐
         ▼                             ▼
  ┌──────────────┐              ┌──────────────┐
  │ Google Cloud │              │  Chrome Ext  │
  │ Backend API  │              │ (WebSocket)  │
  └──────────────┘              └──────────────┘

---

📡 API Reference & Endpoints

Method	Endpoint	Description
POST	/v1/chat/completions	OpenAI-compatible chat & streaming completions (Default: gemini-3.8-flash, Frontier: argon-high)
GET	/v1/models	List all available models, aliases, and thinking levels
POST	/v1/models	Dynamically register custom model hashes at runtime
POST	/v1/models/discover	Autonomously probe Google backend for new unannounced models
GET	/v1/workers	Real-time worker pool health, active accounts & concurrency metrics
GET	/stats	SQLite token accounting, messages count & financial ROI savings
GET	/export/excel	Download comprehensive multi-sheet Excel report (.xlsx)
GET	/v1/analytics	Machine-readable JSON analytics report
GET	/health	Live cookie TTL inspector and cluster health
GET	/help	Terminal developer CLI guide

---

⚙️ CLI Tools

./goapi --help              # Show terminal CLI guide
./goapi --stats             # Print real-time database usage & cost saved
./goapi --export out.xlsx   # Export detailed spreadsheet analytics
./goapi --mcp               # Run as Model Context Protocol (MCP) server

---

🔒 Security & Privacy

• Zero Hardcoded Secrets: Credentials and cookies are held strictly in local files and memory; nothing is sent to external third-party servers.
• Local Network Isolation: All WebSocket discovery and cookie sync happen strictly over localhost or local subnet.
• Safe Version Control: Pre-configured .gitignore prevents private Google session tokens and cookies from ever being committed.

---

🌟 Support & Community

If you find Gemini 4 Argon Free API helpful, please give this repository a ⭐ Star to support ongoing development and help other developers discover Google's frontier reasoning models!

---

📄 License

MIT License. Developed for the developer and AI research community.