A local AI agent can feel slow for several different reasons: the model may take time to load, process a long prompt, generate tokens, wait on CPU or GPU resources, or make repeated calls around tool use. Start by identifying which part is slow before changing hardware. Check model placement, memory pressure, context length, and runtime settings, then measure the same task again.
Find out where the delay happens
Time the agent from the start of a request and note when the delay occurs. A long pause before the first token points to a different problem than slow streaming, a slow tool, or many slow agent steps. These measurements are more useful than a single overall “response time.”
- Before the first token: The model may be loading after idle, processing a large prompt, or reading model files from slow storage.
- Between tokens: Check CPU/GPU placement, memory pressure, and inference settings.
- During a tool call: The tool or its external dependency may be the bottleneck, rather than model generation.
- Between agent steps: Repeated inference and serial tool calls can add substantial end-to-end time even when token generation is acceptable.
For LocalAI, enable debug logging and use a simple streaming request to inspect per-token timing. Keep an end-to-end measurement too: a fast model call does not guarantee a fast agent workflow. LocalAI’s advanced documentation describes its diagnostics and performance guidance.
Check whether the model is actually using the GPU
Do not assume that installing a GPU or choosing a GPU-enabled runtime means the model is running on it. Confirm placement in the runtime’s own status or startup output.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors#1 Best Overall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 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.
Ollama
Run ollama ps while the model is loaded. In the PROCESSOR column, Ollama reports placement such as 100% GPU, 100% CPU, or a split between the two. If it is on CPU when you expected GPU inference, check whether the installed backend and drivers support your hardware, and whether enough VRAM is available for the model and its context. See the Ollama FAQ for placement and configuration details.
llama.cpp
Inspect startup output for messages indicating that layers have been offloaded to the GPU. If the expected offload is absent or limited, check the build, backend, and available GPU memory. The llama.cpp performance tips explain how to interpret and tune performance in its examples.
Serving deployments
In a vLLM serving setup, low GPU utilization does not necessarily mean the GPU itself is the problem. CPU contention in tokenization, scheduling, media loading, or output handling can leave the GPU waiting. Check runtime diagnostics and CPU load alongside GPU utilization. vLLM’s serving optimization guidance focuses on serving behavior and throughput; it is not a guarantee that vLLM will improve a single-user desktop agent.
Rank #2
- 𝗔𝟵 𝗠𝗮𝘅 𝗔𝗜𝟵 𝟰𝟳𝟬 – 𝗙𝗹𝗮𝗴𝘀𝗵𝗶𝗽 𝗔𝗜 & 𝗣𝗿𝗼𝗳𝗲𝘀𝘀𝗶𝗼𝗻𝗮𝗹 𝗪𝗼𝗿𝗸𝘀𝘁𝗮𝘁𝗶𝗼𝗻 - The GEEKOM A9 Max now features the AMD Ryzen AI 9 470, built on AMD’s latest Strix Point architecture. Delivering up to 86 TOPS AI acceleration, including an XDNA 2 NPU rated up to 55 TOPS, this compact mini PC transforms how professionals handle demanding workloads. From running large enterprise AI models and local LLMs to producing 8K video content and advanced 3D rendering, the A9 Max ensures smooth, uninterrupted performance. Perfect for enterprise AI projects, financial analysis, scientific research, professional content creation, educational labs.
- 𝗔𝗔𝗔 𝗚𝗮𝗺𝗶𝗻𝗴 𝗨𝗻𝗹𝗲𝗮𝘀𝗵𝗲𝗱—𝗨𝗽 𝘁𝗼 𝟭𝟯𝟬 𝗙𝗣𝗦 𝘄𝗶𝘁𝗵 𝗜𝗰𝗲𝗕𝗹𝗮𝘀𝘁 𝟯.𝟬 – Powered by AMD Ryzen AI 9 HX 470 (12C/24T, up to 5.2GHz), Radeon 890M Graphics, the GEEKOM A9MAX is built for smooth 1080p AAA gaming, streaming and 4K creation. Radeon 890M platforms have demonstrated up to 90 FPS in Cyberpunk 2077, 99 FPS in Forza Horizon 5 and 130 FPS in F1 24 with optimized settings and supported upscaling or frame generation. The all-metal chassis and IceBlast 3.0 cooling system combine a large copper heatsink, dual heat pipes and a quiet fan, with Standard and Performance modes to help maintain stable performance during long gaming, editing and rendering sessions.
- 𝗛𝗶𝗴𝗵-𝗦𝗽𝗲𝗲𝗱 𝗗𝗗𝗥𝟱 𝗠𝗲𝗺𝗼𝗿𝘆 & 𝗘𝘅𝗽𝗮𝗻𝗱𝗮𝗯𝗹𝗲 𝗦𝘁𝗼𝗿𝗮𝗴𝗲 - Preinstalled with 32GB DDR5 RAM (expandable to 128GB) and equipped with dual PCIe Gen4 NVMe SSD slots (1× M.2 2280 + 1× M.2 2230, up to 8TB total), the A9 Max supports high-capacity storage for large datasets, high-speed scratch disks, and multiple simultaneous workloads. Run AI models, process high-resolution media, or simulate complex projects without delays. This ensures a smooth, responsive, and efficient workflow, enabling professionals to focus on creative and analytical tasks without interruptions.
- 𝟰-𝗗𝗶𝘀𝗽𝗹𝗮𝘆 𝟴𝗞 𝗩𝗶𝘀𝘂𝗮𝗹𝘀 & 𝗗𝘂𝗮𝗹 𝟮.𝟱𝗚𝗯𝗘 𝗡𝗲𝘁𝘄𝗼𝗿𝗸 – Powered by AMD Radeon 890M graphics, GEEKOM A9 Max supports up to four independent displays and 8K output, creating a professional multi-screen workstation without a docking station. Handle financial dashboards, 8K video editing, AI image generation, CAD design, and 3D rendering with ease. Featuring USB4, HDMI 2.1, dual 2.5GbE LAN, WiFi 7, and 3D Stereo WiFi Antenna, it provides stronger signal coverage, fewer dead zones, and more stable wireless connectivity for AI development, creative studios, research labs, and enterprise deployments.
- 𝗨𝗽 𝘁𝗼 𝟱𝟱 𝗧𝗢𝗣𝗦 𝗡𝗣𝗨 𝗳𝗼𝗿 𝗛𝗶𝗴𝗵-𝗖𝗼𝗺𝗽𝘂𝘁𝗲 𝗟𝗼𝗰𝗮𝗹 & 𝗖𝗹𝗼𝘂𝗱 𝗔𝗜 – Combining a 12-core CPU, Radeon 890M graphics and a dedicated NPU, this compact PC supports compatible quantized LLMs and VLMs for batch document intelligence, large-codebase analysis, multi-stream computer vision, generative design and multimodal research. Enterprises can process R&D datasets, proprietary code, financial models and confidential media locally; engineers, developers and creators can accelerate AI prototyping, 8K production, 3D rendering and simulation. Sensitive workloads can remain on-device, while cloud AI adds larger models and deeper reasoning when needed.
Fit the model and context into memory
GPU memory must accommodate both model weights and the key-value (KV) cache used for the active context. If they do not fit, inference may spill work to the CPU or fail to use the desired GPU placement. Long contexts can increase memory use and prompt-processing work, so a larger context window is not automatically faster or better.
Try these changes one at a time, checking that the model still handles the task correctly:
- Use a smaller quantization or a smaller model if its quality is adequate for the task.
- Reduce the context setting to the amount the task needs, and remove irrelevant conversation history.
- Free VRAM used by other processes or reduce the number of GPU layers offloaded.
- Recheck placement and performance after each change; a configuration that saves memory can also change output quality or generation speed.
Ollama’s FAQ currently states a default context window of 4096 tokens, but defaults can change between releases. Check the installed version and set enough context for the actual workload rather than relying on a default or maximizing the window. LocalAI also notes that the prompt plus generated output must fit within the context window. See its performance documentation for related configuration guidance.
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.
Tune CPU threads instead of maximizing them
More threads are not always faster. A high thread count can oversaturate the CPU or compete with other work. In llama.cpp, test thread counts systematically: start low, increase while measuring, and back down if performance worsens or the CPU becomes saturated. LocalAI suggests matching physical cores as a starting point, not a universal rule. Results depend on the runtime and machine.
The llama.cpp performance-tips page includes a measured example using an A6000 with 48 GB of VRAM, a 7-physical-core CPU, 32 GB of RAM, and a specified 30B Q4 model. Its results vary with flags; they should not be treated as a prediction for different hardware, models, or settings.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Reduce avoidable work in the agent workflow
An agent may call the model several times to plan, interpret tool results, and produce an answer. If those calls are sequential, their delays add up. Keep only task-relevant history, ask for output that the next step actually needs, and avoid redundant tool calls. Do not trim context so aggressively that the agent loses information needed to complete the task. Measure total task time as well as individual model-call time.
Rank #4
- Built for Local AI and Advanced Workflows – The BOSGAME M5 AI Mini PC is powered by AMD Ryzen AI Max+ 395 with 16 cores, 32 threads, up to 5.1GHz, 50 TOPS NPU performance and up to 126 TOPS total AI performance. It is designed for local AI inference, private AI assistants, coding, data analysis, virtualization, content creation and demanding multitasking while keeping sensitive data on the device.
- 128GB Unified Memory for Large Models and Creative Projects – M5 includes 128GB LPDDR5X-8000 unified memory, giving the CPU and Radeon 8060S graphics access to a large shared memory pool. This helps support memory-intensive AI workloads, large project files, multiple virtual machines, 3D work, video editing and complex professional applications without the capacity limits of typical 32GB or 64GB mini computers.
- Radeon 8060S Graphics for Creation, Rendering and Gaming – Integrated Radeon 8060S graphics with 40 RDNA 3.5 compute units delivers high-end visual performance without a separate graphics card. Use the M5 creator workstation for 4K video editing, 3D rendering, CAD, AI image workflows, high-resolution media and modern gaming, while maintaining a compact desktop footprint.
- 2TB PCIe 4.0 SSD and Flexible Expansion – A pre-installed 2TB NVMe PCIe 4.0 SSD provides fast access to models, datasets, media libraries and project files. A second M.2 2280 PCIe 4.0 slot allows additional storage expansion, while the SD 4.0 card reader supports efficient photo and video workflows for creators and production teams.
- Professional Connectivity and Four-Display Support – Dual USB4 ports, HDMI 2.1 and DisplayPort 1.4 support up to four displays and resolutions up to 8K@60Hz. WiFi 7, Bluetooth 5.4 and 2.5GbE deliver fast networking for cloud collaboration, NAS access and business deployment. Windows 11 Pro, performance-mode switching, Wake-on-LAN and auto power-on support flexible workstation use.
Avoid unnecessary model reloads
If the first request after a period of inactivity is slow but later requests are quicker, the model may be cold-starting. Ollama documents preloading a model and controlling its residency with keep_alive; its checked FAQ states a default residency period of five minutes. Verify the setting and default for your installed version, since these can change. Keeping a model loaded can reduce reload delays, but it uses memory that may otherwise be available to other models or applications. Consult the Ollama FAQ for the current controls.
Model-file storage matters mainly when loading. LocalAI recommends storing model files on an SSD rather than an HDD, which can help reduce load delays. An SSD does not, by itself, promise faster token generation once the model is loaded. LocalAI’s performance guidance covers this storage recommendation.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Match the fix to the symptom
| What you notice | What to inspect | First useful action |
|---|---|---|
| Long wait before output, especially after idle | Cold start, model loading, storage, or prompt processing | Check timing and logs; try preloading or adjusting model residency. If model files are on an HDD, consider SSD-backed storage for loading. |
| Low generation rate and model shown on CPU | GPU placement, backend or driver support, and VRAM | Check offload output and compatibility; if needed, try a smaller model or supported quantization. |
| CPU and GPU are busy while VRAM is full | Partial offload or memory pressure | Free VRAM, reduce the model or context footprint, or adjust layer offload; then measure again. |
| Performance degrades in long conversations | Context length, KV cache, and prompt-processing load | Remove irrelevant history and use a context setting sized to the task and available memory. |
| GPU utilization is unexpectedly low in a serving setup | CPU-side tokenization, scheduling, media loading, or output processing | Check CPU contention and runtime diagnostics as well as GPU activity. |
| Many slow agent cycles despite acceptable token speed | Repeated model calls and serial tool waits | Remove redundant rounds and pass compact, relevant results between steps; compare end-to-end task time. |
Compare runtimes using your actual workload
Change runtimes only after you have a baseline and know what you need the system to do. Compare the same model, prompt, context, and concurrency on the target machine. Evaluate whether the model and context fit in memory, time to first token, generation rate, output quality and tool-call reliability, hardware and model-format compatibility, and single-user latency versus concurrent-request throughput.
Free tools Windows power users keep installed
One-click scans. No signup required.
Best Value
- LOW ENERGY HIGH PERFORMANCE MINI PC - The Intel Core Ultra 5 125U is part of the Ultra 5 lineup, using the Meteor Lake architecture with BGA 2049. Intel Hyper-Threading technology is available and effectly doubles the core-count of the P-Cores, to a total of 14 threads. Core Ultra 5 125U has 12 MB of L3 cache and operates at 1300 MHz by default, but can boost up to 4.3 GHz, depending on the workload. With a TDP of 15 W, the Core Ultra 5 125U consumes very little energy but outputs high performance efficiency
- 32GB DDR5 RAM + 512GB SSD - The K15 mini computer is equipped with Dual 16GB (Total 32GB) SO-DIMM DDR5 4800MHz memory sticks. 512GB PCIE 4.0 SSD Drive with 3x M.2 2280 Expansion slots. Each slot capable of reading up to 8TB. (24TB MAX)
- QUAD SCREEN 4K DISPLAY SUPPORT - K15 Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and USB Type-C Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support
- OCULINK PORT - The Oculink port on the rear interface enables higher bandwidth capabilities, better frame rates and lower lag. The standard also operates at PCIe x4 speeds, compared to Thunderbolt's x3. Gamers and content creators can benefit from Oculink's higher bandwidth, resulting in better performance and lower lag for eGPU setups
- DUAL NIC FAST 2.5GBE + WIFI 6E + BT 5.2 - Dual Ethernet 2.5GbE LAN port design provides more applications, such as firewall, multichannel aggregation, soft routing, file storage server. Built-in WIFI 6E / Bluetooth 5.2 is more stable and efficient to connect multiple wireless devices such as projector, printer, monitor, speakers and etc
NVIDIA’s Dynamo documentation emphasizes matching software choices to the operating system, model format, GPU architecture and memory, API needs, and throughput target. vLLM describes serving and memory-management optimizations in its optimization guidance. Its 2023 PagedAttention paper reported 2–4× throughput at the same latency against the systems it compared on its evaluated workloads; that serving benchmark is not a promised speedup for a personal local agent. Read the PagedAttention paper.
A runtime that improves concurrent serving throughput may not be the best choice for a single-user setup. Base the decision on the workload you actually run, not a headline benchmark from different hardware or conditions.
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.
