# localLLMrig

> Which hardware do you need to run your LLM locally?

English is the primary language. Dutch mirrors live under /nl. Figures are estimates unless a manufacturer or config URL is attached. Prices always have sourceUrl null. Tokens per second always have benchmarkSourceUrl null. Nothing here is a measured llama-bench result.

## Primary

- [Home](https://localllmrig.com/): budget or model in, three setups out
- [Models](https://localllmrig.com/models): estimated memory per quant at 8192 tokens
- [Hardware](https://localllmrig.com/hardware): PCs, Macs and bare cards, indicative EUR price
- [Methodology](https://localllmrig.com/about): the formula, and which fields have no source

## Machine-readable

- [Models JSON](https://localllmrig.com/api/models): memory at 8192 tokens, estimated true, sourceUrl null
- [Hardware JSON](https://localllmrig.com/api/hardware): indicative EUR prices, sourceUrl null
- [Recommend JSON](https://localllmrig.com/api/recommend?model=llama-3.3-70b&budget=2000&quant=Q4&context=8192): top 3, estimatesAreNotMeasurements true
- [llms.txt](https://localllmrig.com/llms.txt): this file
- [llms-full.txt](https://localllmrig.com/llms-full.txt): formula, catalog and example recommendations

## Models

- [Llama 3.2 3B](https://localllmrig.com/models/llama-3.2-3b): 3.21B, architecture from config. Dutch: https://localllmrig.com/nl/modellen/llama-3.2-3b
- [Llama 3.1 8B](https://localllmrig.com/models/llama-3.1-8b): 8.03B, architecture from config. Dutch: https://localllmrig.com/nl/modellen/llama-3.1-8b
- [Mistral 7B v0.3](https://localllmrig.com/models/mistral-7b-v03): 7.25B, architecture estimated. Dutch: https://localllmrig.com/nl/modellen/mistral-7b-v03
- [Gemma 2 9B](https://localllmrig.com/models/gemma-2-9b): 9.24B, architecture estimated. Dutch: https://localllmrig.com/nl/modellen/gemma-2-9b
- [Qwen2.5 14B](https://localllmrig.com/models/qwen2.5-14b): 14.7B, architecture from config. Dutch: https://localllmrig.com/nl/modellen/qwen2.5-14b
- [Phi-4 14B](https://localllmrig.com/models/phi-4): 14.7B, architecture estimated. Dutch: https://localllmrig.com/nl/modellen/phi-4
- [Gemma 2 27B](https://localllmrig.com/models/gemma-2-27b): 27.2B, architecture estimated. Dutch: https://localllmrig.com/nl/modellen/gemma-2-27b
- [Qwen2.5 32B](https://localllmrig.com/models/qwen2.5-32b): 32.8B, architecture from config. Dutch: https://localllmrig.com/nl/modellen/qwen2.5-32b
- [Llama 3.3 70B](https://localllmrig.com/models/llama-3.3-70b): 70.6B, architecture from config. Dutch: https://localllmrig.com/nl/modellen/llama-3.3-70b
- [Qwen2.5 72B](https://localllmrig.com/models/qwen2.5-72b): 72.7B, architecture from config. Dutch: https://localllmrig.com/nl/modellen/qwen2.5-72b
- [DeepSeek-R1-Distill 70B](https://localllmrig.com/models/deepseek-r1-distill-70b): 70.6B, architecture estimated. Dutch: https://localllmrig.com/nl/modellen/deepseek-r1-distill-70b

## Budgets

- [Local LLM under EUR 800](https://localllmrig.com/budget/800): auto model, Q4, 8192 context
- [Dutch](https://localllmrig.com/nl/budget/800)
- [Local LLM under EUR 1000](https://localllmrig.com/budget/1000): auto model, Q4, 8192 context
- [Dutch](https://localllmrig.com/nl/budget/1000)
- [Local LLM under EUR 1500](https://localllmrig.com/budget/1500): auto model, Q4, 8192 context
- [Dutch](https://localllmrig.com/nl/budget/1500)
- [Local LLM under EUR 2000](https://localllmrig.com/budget/2000): auto model, Q4, 8192 context
- [Dutch](https://localllmrig.com/nl/budget/2000)
- [Local LLM under EUR 3000](https://localllmrig.com/budget/3000): auto model, Q4, 8192 context
- [Dutch](https://localllmrig.com/nl/budget/3000)
- [Local LLM under EUR 5000](https://localllmrig.com/budget/5000): auto model, Q4, 8192 context
- [Dutch](https://localllmrig.com/nl/budget/5000)

## Optional

- [Best GPU for Llama 70B](https://localllmrig.com/best-gpu-for-llama-70b)
- [Best Mac for a local LLM](https://localllmrig.com/best-mac-for-local-llm)
- [Local LLM under EUR 1000, longer page](https://localllmrig.com/local-llm-under-1000)
- [Guides](https://localllmrig.com/guides): English notes. Memory figures are formula estimates. No measured benchmark.
- [Best GPU for local LLMs in 2026](https://localllmrig.com/guides/best-gpu-for-local-llms-2026)
- [How much VRAM does a 70B model need?](https://localllmrig.com/guides/how-much-vram-for-70b)
- [Mac vs PC for a local LLM](https://localllmrig.com/guides/mac-vs-pc-for-local-llms)
- [Used RTX 3090 buying guide (EU)](https://localllmrig.com/guides/used-rtx-3090-buying-guide-eu)
- [Running DeepSeek and Qwen locally](https://localllmrig.com/guides/running-deepseek-and-qwen-locally)
- [Quantization explained for local LLMs](https://localllmrig.com/guides/quantization-explained)
- [The RTX 5090 is (reportedly) dead: best 32GB alternatives for local LLMs](https://localllmrig.com/guides/rtx-5090-reportedly-dead-32gb-alternatives)
- [Is RTX Spark worth it for local LLMs vs Mac Studio vs Strix Halo?](https://localllmrig.com/guides/rtx-spark-vs-mac-studio-vs-strix-halo)
- [Run Nemotron 3.5 Bolt 74B on a 24GB GPU](https://localllmrig.com/guides/nemotron-3-5-bolt-74b-upcoming)
- [Run a local LLM on Windows in 5 minutes (Llama Windows)](https://localllmrig.com/guides/llama-windows-local-llm)
- [Best local LLM rig under $500 / $1000 / $3000](https://localllmrig.com/guides/best-local-llm-rig-under-500-1000-3000)
- [Buy vs rent](https://localllmrig.com/buy-vs-rent): self-hosted purchase divided by a monthly rent you type. No invented cloud rate.
- [2026 Local LLM Build Guide](https://localllmrig.com/build-guide): catalog bands under USD 500, 1000 and 3000. Stripe button hidden until STRIPE_LINK_BUILD_GUIDE is https.
- [Build guide sample](https://localllmrig.com/build-guide/sample)
- [Newsletter](https://localllmrig.com/newsletter): stores a contact only when RESEND_API_KEY and RESEND_AUDIENCE_ID are set; never sends email
- [Affiliate disclosure](https://localllmrig.com/affiliate): shop buttons say Check price and are rel=sponsored only with a real partner id
- [Sponsor](https://localllmrig.com/sponsor): founding sponsor, EUR 590 excl. VAT, 2 slots, 3 months, category exclusivity. Dutch: https://localllmrig.com/nl/adverteren
- [Pricing](https://localllmrig.com/pricing): custom report EUR 49, report plus call EUR 149, early supporter EUR 19. Dutch: https://localllmrig.com/nl/prijzen
- [Sponsors](https://localllmrig.com/sponsors): founding credits. The list is not an input to ranking
- [Privacy](https://localllmrig.com/privacy), [Terms](https://localllmrig.com/terms), [Company details](https://localllmrig.com/company): drafts, company numbers not filled in

## Dutch mirrors

- [Dutch home](https://localllmrig.com/nl)
- [Dutch models](https://localllmrig.com/nl/modellen)
- [Dutch hardware](https://localllmrig.com/nl/hardware)
- [Dutch methodology](https://localllmrig.com/nl/over)