LG Agent Commands — Hermes & OpenClaw

Ollama Cloud Models

Switch models for Hermes & OpenClaw agents

⚡ Switch Model (In Session)

/modelSwitch model for current session

🟠 OpenClaw — Loomy Bot

kimi-k2.6:cloudKimi K2.6 on Ollama Cloud

gemma4:31b-cloudGemma 4 31B on Ollama Cloud

qwen3-coder-next:cloudQwen 3 Coder Next on Ollama Cloud

glm-5.1:cloudGLM 5.1 on Ollama Cloud

minimax-m2.7:cloudMiniMax M2.7 on Ollama Cloud

nemotron-3-super:cloudNemotron 3 Super on Ollama Cloud

💜 Hermes — Soluna & Lumina

kimi-k2.6:cloudKimi K2.6 on Ollama Cloud

gemma4:31b-cloudGemma 4 31B on Ollama Cloud

qwen3-coder-next:cloudQwen 3 Coder Next on Ollama Cloud

glm-5.1:cloudGLM 5.1 on Ollama Cloud

minimax-m2.7:cloudMiniMax M2.7 on Ollama Cloud

nemotron-3-super:cloudNemotron 3 Super on Ollama Cloud

🖥️ Ollama Run (Standalone)

kimi-k2.6:cloudRun Kimi K2.6 directly

gemma4:31b-cloudRun Gemma 4 31B directly

qwen3-coder-next:cloudRun Qwen 3 Coder Next directly

glm-5.1:cloudRun GLM 5.1 directly

minimax-m2.7:cloudRun MiniMax M2.7 directly

nemotron-3-super:cloudRun Nemotron 3 Super directly

🦙 Ollama Locals — Best Models for 16GB Macs

Researched by Loomy Bot. Ranked best to worst. Run local or launch with Hermes.

🥇Qwen3.5 9B5.3 GB

Best overall. Hermes official pick. Beats models 3x its size.

Pullollama pull qwen3.5:9b

Runollama run qwen3.5:9b

Hermeshermes chat -m qwen3.5:9b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/qwen3.5:9b

🥈Qwen3 30B-A3B MoE6 GB

30B quality at 3B speed. The secret weapon for small Macs.

Pullollama pull qwen3:30b-a3b

Runollama run qwen3:30b-a3b

Hermeshermes chat -m qwen3:30b-a3b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/qwen3:30b-a3b

🥉Qwen3-Coder 30B-A3B MoE6 GB

Same MoE trick, optimized for code + agent tasks.

Pullollama pull qwen3-coder:30b-a3b

Runollama run qwen3-coder:30b-a3b

Hermeshermes chat -m qwen3-coder:30b-a3b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/qwen3-coder:30b-a3b

4GLM-4.7-Flash5 GB

Most obedient tool caller. More precise than Qwen.

Pullollama pull glm-4.7-flash

Runollama run glm-4.7-flash

Hermeshermes chat -m glm-4.7-flash --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/glm-4.7-flash

5Qwen3 8B5 GB

Reliable all-rounder with /think mode.

Pullollama pull qwen3:8b

Runollama run qwen3:8b

Hermeshermes chat -m qwen3:8b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/qwen3:8b

6DeepSeek R1 14B8.7 GB

Reasoning champion. Shows thinking chain. Tight fit on 16GB.

Pullollama pull deepseek-r1:14b

Runollama run deepseek-r1:14b

Hermeshermes chat -m deepseek-r1:14b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/deepseek-r1:14b

7Qwen3 14B9 GB

Big quality jump from 8B. Tight on 16GB.

Pullollama pull qwen3:14b

Runollama run qwen3:14b

Hermeshermes chat -m qwen3:14b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/qwen3:14b

8Qwen3-Coder 14B9 GB

Code-specialized 14B. Same tight fit.

Pullollama pull qwen3-coder:14b

Runollama run qwen3-coder:14b

Hermeshermes chat -m qwen3-coder:14b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/qwen3-coder:14b

9Phi-4 14B9 GB

STEM specialist. Fast and precise.

Pullollama pull phi4:14b

Runollama run phi4:14b

Hermeshermes chat -m phi4:14b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/phi4:14b

10Phi-4 Mini 3.8B2.5 GB

Tiny but smart. Good for quick Q&A, NOT for agent tasks.

Pullollama pull phi4-mini

Runollama run phi4-mini

Hermeshermes chat -m phi4-mini --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/phi4-mini

11Llama 3.1 8B4.7 GB

Classic pick, good for RAG/docs.

Pullollama pull llama3.1:8b

Runollama run llama3.1:8b

Hermeshermes chat -m llama3.1:8b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/llama3.1:8b

12Qwen2.5 Coder 7B4.4 GB

Best small dedicated coder.

Pullollama pull qwen2.5-coder:7b

Runollama run qwen2.5-coder:7b

Hermeshermes chat -m qwen2.5-coder:7b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/qwen2.5-coder:7b

13DeepSeek R1 7B5 GB

Basic reasoning on budget.

Pullollama pull deepseek-r1:7b

Runollama run deepseek-r1:7b

Hermeshermes chat -m deepseek-r1:7b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/deepseek-r1:7b

14Llama 3.2 Vision 11B8 GB

Image analysis/OCR. Not for agent work.

Pullollama pull llama3.2-vision:11b

Runollama run llama3.2-vision:11b

Hermeshermes chat -m llama3.2-vision:11b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/llama3.2-vision:11b

15Gemma 3 4B3 GB

Lightweight tasks only.

Pullollama pull gemma3:4b

Runollama run gemma3:4b

Hermeshermes chat -m gemma3:4b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/gemma3:4b

16Llama 3.2 3B2 GB

8GB starter model.

Pullollama pull llama3.2

Runollama run llama3.2

Hermeshermes chat -m llama3.2 --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/llama3.2

17Mistral 7B4.1 GB

Outperformed by Qwen in 2026. Skip.

Pullollama pull mistral:7b

Runollama run mistral:7b

Hermeshermes chat -m mistral:7b --provider custom --base-url http://localhost:11434/v1

Switch/model ollama/mistral:7b