Any Model. One Command.
Hermes Agent doesn't lock you into a single provider. Switch models with hermes model — no config file editing, no code changes.
Nous PortalRecommended
First-party access to Hermes models
nous-hermes-3hermes-3-405b
hermes model
# Select: Nous Portal
OpenRouterMost Flexible
200+ models, one API key
claude-3.5-sonnetgpt-4odeepseek-r1
hermes model
# Select: OpenRouter
OpenAI
Direct OpenAI API
gpt-4ogpt-4o-minio3
hermes model
# Select: OpenAI
Anthropic
Direct Anthropic API
claude-opus-4claude-sonnet-4
hermes model
# Select: Anthropic
Ollama (Local)Privacy-First
Run models on your own hardware
llama3mistralqwen2.5
hermes model
# Select: Ollama
# Enter: http://localhost:11434
Kimi / Moonshot
Moonshot AI's Kimi models
moonshot-v1-128kkimi-latest
hermes model
# Select: Kimi
MiniMax
MiniMax's long-context models
abab6.5s-chatMiniMax-Text-01
hermes model
# Select: MiniMax
z.ai / GLM
Zhipu AI's GLM series
glm-4-plusglm-4-flash
hermes model
# Select: GLM
Switching Models Takes 10 Seconds
$ hermes model
? Select a provider:
❯ Nous Portal
OpenRouter
OpenAI
Anthropic
Ollama
? Select a model: claude-opus-4
✓ Model set to anthropic:claude-opus-4
No restart needed. Change takes effect immediately.
You can also set the model inline: /model openrouter:deepseek/deepseek-r1
Using a Custom or Self-Hosted Endpoint
Any OpenAI-compatible endpoint works with Hermes Agent. This includes LiteLLM proxies, vLLM deployments, Together AI, Fireworks AI, and any provider that follows the OpenAI Chat Completions API spec.
# set via hermes config commands
hermes config set provider openai
hermes config set openai.base_url https://your-proxy.internal/v1
hermes config set openai.api_key sk-your-key
hermes config set openai.model your-model-name
# or via environment variables
export OPENAI_BASE_URL=https://your-proxy.internal/v1
export OPENAI_API_KEY=sk-your-key
Not Sure Which to Use?
| Scenario | Recommended | Reason | Difficulty |
|---|---|---|---|
| Getting started, want quick setup | OpenRouter | One API key covers all major models | Easy |
| Need strongest reasoning | Anthropic / OpenAI | Claude Opus, o3 are best-in-class | Easy |
| Data must stay local | Ollama | Fully local, no network requests | Medium |
| Primarily Chinese tasks | Kimi / GLM | Long context, optimized for Chinese | Easy |
| Want native Hermes models | Nous Portal | Trained specifically for agent tasks | Easy |
| Enterprise internal deployment | Custom Endpoint | Connect your own vLLM or LiteLLM | Medium |
Model FAQs
Does Hermes Agent work with free models?
Yes. Via OpenRouter, you can access several free-tier models including
google/gemini-2.0-flash-exp and meta-llama/llama-3.1-8b-instruct. Free models have rate limits but work for most tasks.Can I use different models for different tasks?
Hermes supports per-task model routing via the
/model slash command mid-conversation, or by configuring the delegate_task tool to spawn subagents on a different model. This lets you use a fast cheap model for routine tasks and a powerful model for complex reasoning.Does switching models lose my memory or skills?
No. Memory and skills are stored independently of the model. Switching from GPT-4o to Claude Opus does not affect your persistent memory, user profile, or installed skills.
Sponsored
Ready to Configure Your Model?
New to Hermes? Start with the install guide →