Models
FOUR MODELS.
TWO PLATFORMS.
Two GPU models on NVIDIA L40S for paid tiers. Two CPU models on IBM POWER10 free for everyone. Every OpenClaw config slot covered — no gaps, no workarounds.
Deep Dive
Qlaw Orchestrator
Starter+Your agent's brain. Plans, reasons, and acts. Powers OpenClaw's orchestrator pattern, coordinates sub-agents, and executes tool calls. Deep thinking mode for complex tasks, direct mode for fast ones. Native tool calling means agents actually complete work, not just describe it.
Agentstool callingreasoningsub-agent coordination
Speed
⚡⚡⚡
Context
64K
Hardware
Qwen3-32B
API ID
qlaw-orchestratorPaid Tiers — NVIDIA L40S GPU
🧠
Qlaw Orchestrator
Your agent's brain
Powered by Qwen3-32B on NVIDIA L40S
Starter+
👁
Qlaw Vision
Your agent's eyes
Powered by Qwen2.5-VL-7B on NVIDIA L40S
Starter+
Free Tier — Power-class servers
💻
Qlaw Coder-Next
Your agent's coder
Powered by Qwen3-Coder-Next on Power-class servers
Free+
⚡
Qlaw Nano
Your agent's reflex
Powered by Qwen3-1.7B on Power-class servers
Free+
| Model | Best For | Speed | Context | Plan |
|---|---|---|---|---|
| Qlaw Orchestrator | Agents, tool calling, reasoning, sub-agent coordination | ⚡⚡⚡ | 64K | Starter+ |
| Qlaw Vision | Screenshots, PDFs, browser views, document analysis | ⚡⚡⚡ | 32K | Starter+ |
| Qlaw Coder-Next | Coding agents, repo analysis, multi-file edits, OpenClaw orchestrator | ⚡ | 100K | Free+ |
| Qlaw Nano | Routing, classification, quick summaries, heartbeat agents, structured extraction | ⚡⚡⚡⚡ | 16K | Free+ |
Model names are role-based and permanent. When the underlying model is upgraded, your config stays the same — only the hardware improves. Paid subscribers vote on all model changes.