Best models for Hermes
Hermes runs your skills locally and leans on the model for planning and skill use. These are the models that pair best with it right now, grouped by what you actually want to spend and ranked on real agentic-coding performance.
As of June 2026, the best model for Hermes comes down to one question: do you have a ChatGPT Codex subscription?
If you do, run two GPT models on your plan instead of paying per token:
If not, route everything through OpenRouter with one API key and pick by need:
- Gemini 3 Flash for fast, cheap, high-volume runs
- DeepSeek V4 for decent quality, even cheaper
- Claude Sonnet 4.6 for great quality, still cheaper than 5.5 and Opus
- Kimi K2.6 for a strong agentic coder at rock-bottom prices
We rank GPT-5.4 first because it is the best value for everyday Hermes work: strong agentic coding at a mid-tier price, on a Codex subscription or the OpenAI API.
GPT-5.5 is the more capable model and tops CursorBench 3.1, scoring 59.2% at medium effort and 64.3% at extra-high, the best result on the board.
A note on cost: GPT-5.5 is about 2x more expensive than GPT-5.4, but it uses tokens much more efficiently, so on a real task it can cost the same or only a little more. Test both on your own workload before you pick one.
For cheaper high-volume work, drop to GPT-5.4 Mini, Gemini 3 Flash, or DeepSeek V4. Reach for Claude Opus 4.8 only when budget is no object.
What most Hermes users should run: strong agentic coding at mid-tier prices, on the OpenAI API or a ChatGPT Codex subscription.
- 11GPT-5.4OpenAIThe best-value default for everyday work. Run it on a Codex subscription or the OpenAI API.Context1.05MInput$2.5/M
- 22GPT-5.4 MiniOpenAICheap and fast for high-volume edits and sub-agents. Codex subscription or OpenAI API.Context400KInput$0.75/M
- 33Claude Sonnet 4.6AnthropicA dependable all-rounder for routine runs. Anthropic API or OpenRouter.Context1MInput$3/M
Cheap, fast models for high-volume automation and routine edits, at a fraction of flagship prices.
- 11Gemini 3 FlashGoogle DeepMindThe fastest cheap pick for high-volume runs. Route it through OpenRouter.Context1.049MInput$0.5/M
- 22DeepSeek V4DeepSeekThe cheapest pick here, and open-source. Route it through OpenRouter.Context1.049MInput$0.435/M
The most reliable model for hard, multi-step planning. Worth it for high-stakes runs, overkill for everyday work.
- 11Claude Opus 4.8AnthropicThe top pick when budget is no object. Anthropic API only.Context1MInput$5/M
Each model's Artificial Analysis Intelligence Index score against its blended price per 1M tokens. Toward the top right is more intelligence per dollar.
What is the best model for Hermes?
For most people, GPT-5.4 is the best model for Hermes: the best value for everyday agentic coding. GPT-5.5 is more capable and tops CursorBench 3.1 (59.2% at medium effort, 64.3% at extra-high), but it uses tokens efficiently enough that the real cost gap is small, so test both on your own workload. Claude Sonnet 4.6 is the dependable Anthropic-API pick.
What is the cheapest good model for Hermes?
GPT-5.4 Mini at $0.75 per million input tokens is the cheapest GPT that still handles real agent work. Gemini 3 Flash at $0.50 per million is the fastest cheap pick, and DeepSeek V4 is the cheapest overall at $0.435 per million, with open weights if you want them.
How do you actually run each model with Hermes?
GPT models (5.4, 5.4 Mini, and 5.5) run on a ChatGPT Codex subscription or the OpenAI API. Claude models (Sonnet 4.6 and Opus 4.8) are Anthropic API only. Everything else, including Gemini 3 Flash and DeepSeek V4, routes through OpenRouter with a single key, the simplest way to switch between non-OpenAI models.
Do you need Claude Opus 4.8 for Hermes?
No. Opus 4.8 is the most capable model for multi-step planning, but it is the most expensive and overkill for everyday Hermes work. Most people are better served by GPT-5.4 or GPT-5.5, dropping to GPT-5.4 Mini, Gemini 3 Flash, or DeepSeek V4 for high-volume runs. Reach for Opus only when budget is no object.
Best cheap models for Hermes
Best free models for Hermes
Best open-source models for Hermes
Agent
HermesModels
7Updated
June 2026