Free Models for OpenClaw: What Actually Costs Nothing (2026)

Model names and free-tier availability checked on October 14, 2026. Check Google’s current pricing and your live project quota before depending on free cloud access.

“Free” gets thrown around a lot in the AI model space. Let me be specific about what that actually means for OpenClaw users.

There are three categories: models that cost literally nothing (local), models with free tiers that eventually run out, and models so cheap they round to zero on most invoices. I’ll cover all three.

Actually free: local models

The only models that cost nothing per-token are the ones running on your own hardware. Ollama makes this straightforward.

Install it, pull a model, point OpenClaw at it:

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a coding model
ollama pull qwen3.8:27b

Then add Ollama as a provider in ~/.openclaw/openclaw.json:

{
  models: {
    providers: {
      ollama: {
        baseUrl: "http://localhost:11434/v1",
        api: "openai-completions",
        models: [
          { id: "qwen3.8:27b", name: "Qwen3.8 27B" }
        ]
      }
    }
  },
  agents: {
    defaults: {
      model: { primary: "ollama/qwen3.8:27b" }
    }
  }
}

Which local models work well

Qwen3.8 27B is a newer coding and agent candidate. Ollama’s listing shows an approximately 18GB download and tools support. Leave additional memory for context and runtime buffers; a 24GB GPU or 32GB+ unified-memory machine is a more practical starting point than a 16GB machine.

Qwen3 Coder 30B remains a coding-focused alternative. Muse Glimmer 30B is another local agent candidate. Compare them on the same tool sequence rather than treating release order as a quality ranking.

gpt-oss 20B, Gemma 4 E4B, and Qwen3 8B are options for smaller machines. Fit depends on the exact quantization and context size. Keep a fallback for tasks the local model cannot finish.

The honest tradeoff

Local models are free but not fast. Response times are 3-10x slower than cloud APIs, depending on your hardware. And even the best consumer-sized open models still trail frontier Claude models on the hardest debugging and multi-file work.

If your work involves mostly reading files, generating boilerplate, and simple edits, local works well. If you’re doing complex debugging or architectural work, you’ll want a cloud model for those tasks.

Free tiers from cloud providers

A few providers offer genuinely free usage up to a limit.

Gemini 3.8 Flash and Gemini 3.5 Flash-Lite have free-tier entries in Google’s current pricing. Use Flash for the broader coding comparison and test Flash-Lite on lighter work. Limits vary by model and project; inspect your AI Studio quota instead of copying a requests-per-minute number from an old article. Gemini 4 Argon is paid-only.

// Add to ~/.openclaw/openclaw.json
{
  models: {
    providers: {
      google: {
        models: [
          { id: "gemini-3.8-flash", name: "Gemini 3.8 Flash" }
        ]
      }
    }
  }
}

The catch: the free tier has stricter rate limits and your data may be used for training. For side projects and learning, that’s probably fine. For proprietary code, use the paid API or run local.

Almost free: sub-dollar models

Budget cloud models are paid, even when their token rates are low. Current candidates include:

Model	haimaker model ID	Workload to test
DeepSeek V4.1 Flash	deepseek/deepseek-v4.1-flash	Routine agent work
GLM-5.3 Flash	z-ai/glm-5.3-flash	Budget tool use and summaries
GPT-6 Luna	openai/gpt-6-luna	Focused, high-volume tasks
Claude Haiku 5.5	anthropic/claude-haiku-5-5	Lightweight cloud work
MiniMax M3	minimax/minimax-m3	Coding and agent workflows

Check the live catalog for input and output rates. Estimate your actual token volume, retries, and tool charges before promising a daily or monthly budget.

You can access all of these through Haimaker with a single API key, or set them up individually with each provider.

Once your Haimaker key is in HAIMAKER_API_KEY, the connect CLI (npx -y @haimaker/connect --openclaw) drops the provider config into OpenClaw for you, so you can route those sub-dollar models without touching JSON.

The practical setup: hybrid free + cheap

Most people who care about costs end up here:

1. Local model for simple tasks (Qwen3.8 27B via Ollama, free)
2. Cheap cloud model for medium tasks (DeepSeek V4.1 Flash or MiniMax M3, pennies)
3. Premium model for hard problems (Claude Sonnet or Opus, pay-per-use)

{
  agents: {
    defaults: {
      model: {
        primary: "ollama/qwen3.8:27b",
        thinking: "anthropic/claude-sonnet-5-5"
      }
    }
  }
}

Use the local model for tasks it completes reliably, then switch to a stronger cloud model for failures. Measure that split on your own workload; it does not guarantee a fixed daily bill.

For more on model routing, see our guide on multi-agent workflows.

GET $10 FREE CREDITS ON HAIMAKER

---

For a full model comparison, see best models for OpenClaw. For cost optimization strategies, see cutting token costs by 96%.