Jatevo

Jatevo — Free LLM Inference for OpenClaw

Jatevo.id provides free inference-as-a-service for open-source LLMs. Zero cost, production-grade, hosted in Asia with international endpoints.

Why Jatevo?

• $0 cost — All models are completely free
• 3 flagship models — Qwen 3.5 Plus, Kimi K2.5, GLM 4.7
• Blazing fast — GLM 4.7 at 448 tokens/sec, Qwen 3.5+ at 50 tok/s
• OpenAI-compatible — Drop-in replacement, works with any OpenAI SDK
• 1M token context — Qwen 3.5 Plus supports up to 1 million tokens
• Production-grade — 32 concurrent requests tested, fully parallel

Quick Setup (30 seconds)

Option 1: Run the setup script (recommended)

bash setup.sh

This will:

1. Ask for your Jatevo API key (get one free at jatevo.id)
2. Patch your OpenClaw config with all available models
3. Optionally set Jatevo as your default model
4. Test the connection

Option 2: Manual setup

1. Get your API key at jatevo.id — sign up and generate a key (starts with jk_)

2. Add to your OpenClaw config (~/.openclaw/openclaw.json):

Add this under models.providers:

{
  "models": {
    "mode": "merge",
    "providers": {
      "jatevo": {
        "baseUrl": "https://jatevo.id/api/open/v1/inference",
        "apiKey": "jk_your_key_here",
        "api": "openai-completions",
        "models": [
          {
            "id": "qwen3.5-plus",
            "name": "Qwen 3.5 Plus",
            "reasoning": true,
            "input": ["text", "image"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 1000000,
            "maxTokens": 65536
          },
          {
            "id": "kimi-k2.5",
            "name": "Kimi K2.5",
            "reasoning": true,
            "input": ["text", "image"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 262144,
            "maxTokens": 32768
          },
          {
            "id": "glm-4.7",
            "name": "GLM 4.7",
            "reasoning": true,
            "input": ["text"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 128000,
            "maxTokens": 40000
          }
        ]
      }
    }
  }
}

3. Set your API key as an environment variable:

## Replace "jk_your_key_here" above with your actual Jatevo API key
## The key is embedded directly in openclaw.json — no .env needed

4. Set as default model (optional):

{
  "agents": {
    "defaults": {
      "model": { "primary": "jatevo/qwen3.5-plus" }
    }
  }
}

5. Restart the gateway:

openclaw gateway restart

Available Models

Qwen 3.5 Plus — jatevo/qwen3.5-plus

The best all-rounder. Handles everything from general conversation to complex analysis.

Performance: Benchmarked at ~95% of Claude Opus 4.6 quality. Handles 1M context effortlessly — feed it entire codebases, books, or document collections. Vision capability lets it analyze images, charts, and screenshots.

Speed breakdown (tested March 2026):

• Short prompts: 38 tok/s
• Code generation: 65 tok/s
• Long-form writing: 41 tok/s
• Average across tasks: 50.7 tok/s

Kimi K2.5 — jatevo/kimi-k2.5

Moonshot AI's reasoning powerhouse. Excels at step-by-step logic and complex problem solving.

Performance: Kimi K2.5 shines on tasks requiring chain-of-thought reasoning. Particularly strong at math, logic puzzles, and multi-step coding problems. Vision support for analyzing diagrams and technical drawings.

Speed breakdown (tested March 2026):

• Short prompts: 29 tok/s
• Code generation: 42 tok/s
• Long-form writing: 41 tok/s
• Average across tasks: 39.5 tok/s

GLM 4.7 — jatevo/glm-4.7

The speed demon. Zhipu AI's GLM family delivers the fastest inference of any model in this tier.

Performance: GLM 4.7 is 9x faster than Qwen 3.5 Plus and 11x faster than Kimi K2.5. Best bilingual model for Chinese and English content. Ideal for latency-sensitive applications, chatbots, and high-throughput batch processing.

Speed breakdown (tested March 2026):

• Short prompts: 327 tok/s
• Code generation: 672 tok/s 🔥
• Long-form writing: 375 tok/s
• Average across tasks: 448.3 tok/s

Model Comparison

When to Use Which

• "I need the smartest model" → jatevo/qwen3.5-plus
• "I need deep reasoning" → jatevo/kimi-k2.5
• "I need it NOW" → jatevo/glm-4.7 (448 tok/s is near-instant)
• "I have a huge document" → jatevo/qwen3.5-plus (1M context)
• "I need to analyze an image" → jatevo/qwen3.5-plus or jatevo/kimi-k2.5
• "I'm building a chatbot" → jatevo/glm-4.7 (speed matters for UX)

Aliases

Add these to your OpenClaw config for quick model switching:

{
  "agents": {
    "defaults": {
      "models": {
        "jatevo/qwen3.5-plus": { "alias": "qwen" },
        "jatevo/kimi-k2.5": { "alias": "kimi" },
        "jatevo/glm-4.7": { "alias": "glm" }
      }
    }
  }
}

Then switch models with /model qwen, /model kimi, or /model glm.

Rate Limits

• 1,200 prompts per 5 hours (4 req/min sustained)
• Fully parallel — tested 32 concurrent requests successfully
• No additional rate limits from Jatevo

Troubleshooting

• "Invalid API key" → Make sure your key starts with jk_
• Rate limited → Wait a few minutes, the window resets on a rolling basis
• Model not found → Run openclaw models list to verify the provider loaded
• Gateway didn't pick up changes → Run openclaw gateway restart

Support

• Website: jatevo.id
• Questions: Open an issue or ask in the OpenClaw Discord

Source: lucacadalora/jatevo-openclaw — distributed by TomeVault.