RE-TECH / LOCAL AI WORKSTATIONS
Ready configurations for local language models, personal assistants and document workflows. Choose by graphics memory — it decides which models you can run. Every part can be changed in the configurator.
Reserve any build with a refundable €10 deposit or send a free enquiry. New parts with manufacturer warranty, serviced graphics cards with a 6-month Re-Tech warranty. Ready 3–5 working days after the parts arrive.
LOCAL AI
Choose the model, software and access to your system.
When the entire workflow is local, without external integrations.
Optional setup: Ollama, LM Studio and models installed and tested before handover.
01 / CONFIGURATIONS
Current parts and prices, straight from our catalogue. Model sizes are estimates for models fully in video memory; real speed depends on the model, quantisation and context length.
Модели до ~14B — първа стъпка в локалния AI
€1.300,00Assembly, stress test and warranty included
We could not add the deposit to the cart. Try again or send an enquiry.
Модели до ~32B, RAG и автоматизация
€4.200,00Assembly, stress test and warranty included
We could not add the deposit to the cart. Try again or send an enquiry.
RTX 4090 — AI, 4K видеомонтаж и 3D
€6.650,00Assembly, stress test and warranty included
We could not add the deposit to the cart. Try again or send an enquiry.
More VRAM = bigger models, not more speed. Two builds with the same VRAM run the same model sizes; the stronger card runs them faster and handles longer context. For models above ~32B you need two graphics cards — see below.
02 / DUAL GPU
Large models are split across both cards. We confirm the PCIe lanes, motherboard, case clearance, power supply and cooling for every dual build.
2× RTX 3090 — модели до ~70B и GPU рендеринг
€7.700,00Assembly, stress test and warranty included
We could not add the deposit to the cart. Try again or send an enquiry.
24GB total VRAM · 2 × 12GB — an affordable way to run ~32B models.
Price on request
Ask for a quote →The memory is split across two cards, not one shared pool. Using both depends on the model and software, and two cards do not double the speed.
03 / LOCAL LLM
An LLM is a language model that processes text and generates answers. Instead of sending every request to a cloud service, you run a compatible model on your own machine.
We consider your tasks, language, licence and memory. A bigger model is not always the better fit.
Software such as Ollama or LM Studio loads the model. Graphics memory decides what fits; the card decides how fast it answers.
Chat, coding help or working with your documents. RAG feeds retrieved information to the model — it does not train a new model.
Hardware and software setup are separate. Add "Local AI — ready to use" in the configurator and we install Ollama, LM Studio and models and test them before handover. RAG, integrations and team setups are quoted separately.
04 / RE-TECH AI
Tell us the task and the budget — we recommend a build, confirm the parts, warranty and lead time in a quote. No commitment.
Choose a starting point. In the next step you add your name, phone, email and details in our protected form.
05 / FAQ
Locally installed software and a downloaded model work offline. Downloads, updates, web search and external services need an internet connection.
Only in a fully local workflow. Cloud models, external APIs and other integrations can send data outside the machine. Access and security settings are planned separately.
Each card shows the largest model size that fits in its graphics memory. Speed depends on the exact model, quantisation, context and software — open the build in the configurator to see the estimate per card, or tell us your target model and we check it.
The focus is running existing models. Fine-tuning is a separate workload assessed individually; it is not the same as training a large model from scratch.
All parts are new with manufacturer warranty. Only the graphics cards are serviced by Re-Tech, tested and covered by a 6-month warranty. Every PC goes through a stress test with a written report.
Reserve with a €10 deposit (deducted from the price, fully refunded if you decline the offer) or send a free enquiry. We confirm the parts and the final price in a quote within 1 working day for reservations. The PC is ready 3–5 working days after the parts arrive.
