You can run Hermes Agent with a model served by Ollama on your own machine by connecting Hermes to Ollama’s local OpenAI-compatible API. The connection is straightforward; the harder parts are choosing a model that can call tools, serving enough context for agent workflows, and fitting the model into your available memory. A local inference setup also does not make optional web, messaging, or cloud-fallback services local.
How Hermes connects to Ollama
Ollama serves models through a local API, and Hermes can treat that API as a custom OpenAI-compatible provider. The Hermes setup guide uses http://localhost:11434/v1; Ollama’s integration guide uses the equivalent loopback address http://127.0.0.1:11434/v1. For a same-machine setup, choose one URL and use it consistently. A local Ollama connection does not require an API key.
As an Amazon Associate I earn from qualifying purchases.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
There are two setup routes: let Ollama guide the installation and configuration, or install Ollama and configure Hermes yourself. Both ultimately point Hermes at the model name Ollama serves. The model examples in documentation can change as catalogs and templates evolve, so check current support and your own model’s behavior rather than relying on a name alone.
Choose a model that can actually use tools
Conversational ability is not enough for an agentic workflow. Hermes needs the model to produce tool calls for actions such as using the terminal or working with files. A model may respond coherently to chat while failing to invoke those actions. Hermes’ guide states: “Models without tool-call support can only chat; they can’t take actions.” See the Hermes Ollama setup guide and Ollama’s Hermes integration page for their current examples.
#1 Best Overall
- EVOLUTION AMD RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
The Hermes guide lists Gemma 4 31B as an example with tool calling and identifies Gemma 2 27B, Gemma 2 9B, and Llama 3.2 3B as examples without tool calling for the tasks shown in that guide. Those examples describe the documentation’s tested or presented support, not a permanent ranking of every model version or template. Ollama’s integration documentation currently names Gemma 4 and Qwen 3.6 as local options. Before settling on a model, verify its current tool-call support and test it with the tools you intend Hermes to use.
- Tool behavior: Can it return usable tool calls for your intended Hermes actions, not just explain what it would do?
- Served context: Can this model and runtime provide the context length your workflow needs?
- Memory and storage: Does the chosen model and quantization fit your RAM or VRAM, and is there room on disk?
- Speed: Is generation and first-turn prompt processing acceptable on your hardware?
- Data boundaries: Does the complete workflow avoid external tools and cloud fallback if that is a requirement?
Check context length before setup
Context length is a functional requirement, not just a model specification to glance at. Hermes’ Ollama guide says agentic work with tools requires at least 64,000 tokens, while its documented Ollama default context is 2,048 tokens. The model’s advertised context capacity does not by itself prove that the running Ollama configuration is serving that much context. Configure and verify the actual runtime setting against the workflow you plan to use. The figures are from the live Hermes documentation accessed October 5, 2026, and are subject to change.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Long prompts also affect responsiveness. Hermes sends its system prompt and schemas for enabled tools with API calls. On CPU-only or low-VRAM systems, processing this input (prefill) can leave the first response silent for minutes; Hermes describes this as expected behavior rather than necessarily a hang. Keeping the model loaded, increasing Hermes’ timeout, checking prompt size, and disabling unused toolsets are the guide’s suggested ways to address that delay. See Hermes’ Ollama guide for the current configuration details.
Rank #2
- Built for Local AI Development: AMD Ryzen AI Halo is designed for local AI development and inference, featuring 128GB unified memory and support for up to 200B parameter models to build and run intensive AI workloads locally.
- 128GB Unified Memory: Features 128GB LPDDR5x unified memory at 8000 MT/s with 256 GB/s memory bandwidth, providing a shared memory pool across the CPU, GPU, and NPU to support larger AI models.
- AMD Ryzen AI Max+ 395 Processor: Features 16 cores, 32 threads, and Zen 5 architecture, paired with AMD Radeon 8060S integrated graphics featuring 40 RDNA 3.5 compute units and an AMD XDNA 2 NPU with up to 50 TOPS.
- Linux AI Developer Platform: Purpose-built for Linux-based AI development with full AMD ROCm software support and preloaded tools, models, and workflows optimized for local AI development.
- Compact, Connected Design: Includes a 2TB M.2 SSD, 10GbE LAN, Wi-Fi 7, Bluetooth 5.4, USB-C connectivity, and HDMI 2.1b.
Install and connect Hermes
Guided setup with Ollama
Ollama documents ollama launch hermes as its quick-start route. It can prompt to install Hermes, let you choose a model, configure the local Ollama endpoint, and offer messaging-gateway setup. Prompts and model choices may change; read each prompt carefully, especially if you want to keep the workflow limited to local inference.
Manual setup
- Install and start Ollama. Check the installation with
ollama --version. You can inspect the local model-tags endpoint in a browser or withcurl http://localhost:11434/api/tagsto see models Ollama reports as available. - Pull a suitable model. Use an Ollama model that fits your hardware and has current tool-call support for the intended actions. For example, replace
MODEL_NAMEwith the exact Ollama model identifier you selected:ollama pull MODEL_NAME. - Point Hermes at Ollama. Run
hermes setupand select a custom endpoint/provider, then enterhttp://localhost:11434/v1, the model identifier, and no API key. Alternatively, set the equivalent fields in~/.hermes/config.yaml:
model:
provider: "custom"
default: "MODEL_NAME"
base_url: "http://localhost:11434/v1"- Start Hermes and test an action. Give it a harmless, bounded task that requires a tool, such as listing a directory you choose. Confirm the action occurred as intended. A fluent text reply alone does not verify tool calling.
Ollama’s integration page documents the alternative loopback spelling http://127.0.0.1:11434/v1. Use either loopback address for a service on the same machine; do not substitute a remote host unless you intentionally serve Ollama elsewhere and understand the network exposure.
Estimate hardware needs and speed
Hermes documentation offers the following starting guidance, not a compatibility guarantee. Actual fit and speed vary with the model, quantization, context length, runtime configuration, and workload. The figures below are those published in the live Hermes documentation accessed October 5, 2026.
Rank #3
- EVOLUTION AMD RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 64GB pool, which is perfect for running LLMs such as Deepseek 32B, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 4% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
| Resource or workload | Hermes guide figure | How to interpret it |
|---|---|---|
| System memory | 8 GB RAM minimum guidance for 3B models; 32+ GB recommended for 27B+ models | Model size alone does not determine fit; quantization, context, and other running processes also use memory. |
| Free storage | 5 GB minimum guidance; 30+ GB recommended for multiple models | Downloaded models take disk space, and keeping several variants increases the requirement. |
| CPU | 4 cores minimum guidance; 8+ cores recommended | More cores do not guarantee a particular tokens-per-second rate. |
| GPU | NVIDIA GPU with 8+ GB VRAM recommended, not required | A GPU is an optional acceleration path, not a prerequisite for running locally. |
| CPU generation example | About 10 tokens/second for a 9B model on a modern 8-core CPU; about 2–5 tokens/second for a 31B model on CPU, with an example response taking 30–120 seconds | These are illustrative estimates in the Hermes guide, not independently measured benchmarks or performance promises. |
| Partial GPU offload example | A 31B model on a 12 GB GPU may partially offload about 40 layers | This is an example only; actual offload depends on the model, quantization, context, and runtime. |
CPU-only use is possible, but larger models may feel slow compared with hosted inference, and long-context prefill can make the first turn especially delayed. If the machine begins swapping to disk under memory pressure, the Hermes guide recommends using a smaller model or adding memory. For a graphics card, compare its usable VRAM with the specific model and context you plan to serve; the guide does not establish a universal best card.
Keep the model available and diagnose failures
- Model unloads after idle time: Hermes documentation says Ollama unloads models after five minutes idle by default. Set a longer keep-alive if repeated reloads are disrupting your workflow, and check
ollama psto see what is currently running and whether GPU layers are offloaded. - First response appears stuck: On CPU-only or low-VRAM hardware, tool schemas and the system prompt can take a long time to prefill. Allow for that processing, consider a longer Hermes timeout, measure prompt size, and disable toolsets you do not use.
- Responses are unexpectedly slow: Check available memory and whether the system is swapping. Try a smaller model or reduce competing memory use; verify GPU offload with
ollama psrather than assuming the GPU is engaged. - Hermes reports no endpoint configured: Set the Ollama base URL in Hermes’ custom-provider configuration. Confirm that Ollama is running and that the configured URL matches the local API endpoint.
- The model chats but does not act: Test tool-call behavior directly and choose a model that supports the intended tools. A successful connection to Ollama is not proof that the model can execute an agent workflow.
Configuration labels and runtime behavior can evolve. Hermes’ provider documentation and local-model guide provide broader context for provider and local-model choices.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Know which parts of the workflow may use the network
With the configuration above, model inference is served by Ollama on the local machine. That fact alone does not establish that every part of a Hermes workflow stays offline. Hermes also documents web browsing, Telegram and Discord gateways, and cloud fallback providers. Those optional tools or providers may communicate with external services, and their use can involve sending requests or data beyond the local inference endpoint.
Recommended Free Tools
Rank #4
- 【Leading AI Mini Workstation】MINISFORUM AI MS-S1 Max Workstation comes with AMD Ryzen AI Max+ 395 processor, which uses AMD's latest generation Zen 5 architecture. It has 16 Cores and 32 Threads, the boost clock is up to 5.1GHz. The overall processor performance is up to 126 TOPS, and the NPU performance reaches up to 50 TOPS. AMD Ryzen AI enables improved productivity, advanced collaboration, and improved efficiency.
- 【AMD Radeon 8060S Graphics 】The MS-S1 Max Mini PC equipped with AMD Radeon 8060S Graphics which built on the new generation of RDNA 3.5 architecture AMD graphics, it brings ultra-high frame rate experiences and advanced content creation features anywhere and delivers staggering performance. It can handle all your computing and multimedia tasks efficiently.
- 【Five 8K Video Output】This MS-S1 Max Workstation comes with five video outputs, 1x HDMI (8K@60Hz), 2x USB4(40Gbps,Alt DP2.0,PD out 15W) and 2x USB4 V2(80Gbps,Alt DP2.0,PD out 15W) Outputs, which support multiple monitors display at the same time and provide a larger and wider filed of view and improve your work efficiency. It is used in fields that require high-performance computing and graphics processing, including digital signage and securities trading, as well as work that uses CAD, such as engineering design, scientific calculations, animation production, and post-production for movies and television
- 【 Fast and Stable Wire & Wireless Speed】It comes with Two 10G Lan Ports for wired connection and and Wi-Fi 7 / BT5.4 for wireless connection, which increased the network speed greatly and expand its functions and improved performance of computer to a large extent and allows you to use more networks such as software routers (OpenWRT / DD-WRT / Tomato etc.), firewalls, NAT, network isolation etc.
- 【Large Storage & Flexible Expandability】This Workstation equipped with 64GB LPDDR5-8000MHz + 2TB M.2 2280 PCIe4.0 SSD. There is another PCIe4.0 SSD slot available for up to 8TB, these SSD slots are compatible with RAID0 and RAID1, you can store movies, videos, photos, important files easily. What’s more, it also comes with 1x standard PCIex16 slot(PCIe4.0x4) inside.
If offline operation is a requirement, do not configure cloud fallbacks, and leave out network-facing tools and messaging integrations. Review which tools are enabled and what each one contacts before using sensitive data. “Local model” describes where inference happens; the privacy boundary for the entire workflow depends on the other integrations you enable.
Running inference locally avoids sending prompts to a hosted model provider for that inference, but it is not a claim that the total workflow has no costs. Hardware, electricity, and any optional cloud service can still carry costs.
When to choose a different local route
Ollama is one way to serve a local model to Hermes, not the only one. Hermes Desktop also offers a managed local-model path using a llama.cpp runtime; it is distinct from configuring Hermes to use Ollama. Hermes documentation additionally covers cloud and other providers. The practical choice depends on whether you want Ollama’s local model-serving workflow, a managed desktop runtime, or a hosted provider, and on your requirements for hardware control, convenience, speed, and data handling. The Hermes local-model guide describes its managed local option.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minute
