New and updated guides on running AI models locally, and every change to our estimates.https://localllmchecker.com/https://localllmchecker.com/guides/how-much-vram-do-i-need/https://localllmchecker.com/guides/how-much-vram-do-i-need/An 8B model at 4-bit needs about 6 GB, not 8. The formula behind that number, and why it moves with the model and the context window.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/memory-bandwidth-and-tokens-per-second/https://localllmchecker.com/guides/memory-bandwidth-and-tokens-per-second/Cards with the same VRAM can answer at very different speeds. The formula that explains it, checked against a real RTX 3050 run.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/quantization-explained/https://localllmchecker.com/guides/quantization-explained/What Q4_K_M and IQ3_XXS actually mean, what each level costs in quality, and how to choose between a big model squeezed small and a small one kept intact.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/kv-cache-explained/https://localllmchecker.com/guides/kv-cache-explained/The KV cache is the memory nobody budgets for. What it holds, why it grows with every token, and the four designs that keep it small.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/gguf-awq-gptq-formats-explained/https://localllmchecker.com/guides/gguf-awq-gptq-formats-explained/Model files come in half a dozen formats and they are not interchangeable. What each one is for, which runtime reads it, and which to pick.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/llama-cpp-offload-flags-explained/https://localllmchecker.com/guides/llama-cpp-offload-flags-explained/The handful of llama.cpp options that move VRAM and tokens per second, what each one trades, and a sensible starting command for a small GPU.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/mixture-of-experts-vram-explained/https://localllmchecker.com/guides/mixture-of-experts-vram-explained/A 30B MoE model needs the memory of a 30B and generates like a 3B. Why, when that is a bargain, and how to run a 120B on a mid-range GPU.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/local-llm-glossary/https://localllmchecker.com/guides/local-llm-glossary/Plain definitions of the terms you hit when you run models yourself, from VRAM and tokens per second to prefill, offload and mixture of experts.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/ollama-context-length-and-vram/https://localllmchecker.com/guides/ollama-context-length-and-vram/Ollama quietly picks a context window and a GPU split for you. What it chooses, how to check it with ollama ps, and what forcing it did on our RTX 3050.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/best-gpu-for-local-llm/https://localllmchecker.com/guides/best-gpu-for-local-llm/Which graphics card to buy for running models at home, ranked by memory and bandwidth rather than by gaming scores, with the strongest model each one runs.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/cpu-and-system-ram-for-local-llm/https://localllmchecker.com/guides/cpu-and-system-ram-for-local-llm/Memory speed and channel count set how fast a CPU generates tokens, and the core count barely does. Real numbers from a DDR4 machine and a table for faster RAM.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/local-llm-power-and-running-cost/https://localllmchecker.com/guides/local-llm-power-and-running-cost/A 350 W GPU costs pennies an hour to run a model, and the bill that surprises people is the idle one. A small calculator and the numbers behind it.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/mac-vs-pc-for-local-llm/https://localllmchecker.com/guides/mac-vs-pc-for-local-llm/A Mac can hold a 70B model no single GPU can, and a GPU answers faster on anything that fits. Where each wins, with speeds from our calculator.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/multi-gpu-local-llm/https://localllmchecker.com/guides/multi-gpu-local-llm/A second graphics card adds memory, not speed. How llama.cpp splits a model across cards, when two cheap cards beat one expensive one, and when they do not.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/used-gpus-for-local-ai/https://localllmchecker.com/guides/used-gpus-for-local-ai/A used RTX 3090 is the best memory per dollar for local models, and also where buyers get burned. Which cards make sense, which are traps, how to test one.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/what-can-you-run-with-12gb-vram/https://localllmchecker.com/guides/what-can-you-run-with-12gb-vram/12 GB is the first tier where local models feel roomy: 14B models at Q5, a 20B mixture-of-experts at the edge, and a used card that doubles the speed.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/what-can-you-run-with-16gb-vram/https://localllmchecker.com/guides/what-can-you-run-with-16gb-vram/16 GB loads 24B to 27B models at 4 bits and 14B models at Q8 with a long window. Which pairing to pick, and why one 16 GB card can be 3 times faster than another.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/what-can-you-run-with-24gb-vram/https://localllmchecker.com/guides/what-can-you-run-with-24gb-vram/24 GB is the tier where local models stop being a compromise. 27B to 32B models at Q4 to Q6, a fast MoE at long context, and what a 70B really costs on one card.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/what-can-you-run-with-32gb-vram/https://localllmchecker.com/guides/what-can-you-run-with-32gb-vram/32 GB holds a 30B model at Q6 and a 27B at Q8 with a long window. Where it falls short of 70B, and what a Mac with 32 GB really offers.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/what-can-you-run-with-48gb-vram/https://localllmchecker.com/guides/what-can-you-run-with-48gb-vram/48 GB holds a 70B model at 4 bits with a usable window. One workstation card or two used 3090s, what context costs, and which large MoE models fit.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/what-can-you-run-with-6gb-vram/https://localllmchecker.com/guides/what-can-you-run-with-6gb-vram/A 6 GB card runs 7B to 9B models with the right quantization and a short window. What fits, what to avoid, and how it held up on a real RTX 3050.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/what-can-you-run-with-8gb-vram/https://localllmchecker.com/guides/what-can-you-run-with-8gb-vram/An 8 GB card is the most common setup and the first to get cramped. Which 7B to 9B models run well, where long context breaks, and why 8 GB cards differ 3x.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/flux-sdxl-low-vram/https://localllmchecker.com/guides/flux-sdxl-low-vram/SDXL runs on 6 GB and Flux on 12 GB with the right precision. The diffusers, ComfyUI and A1111 settings that make each fit, and what each one costs you.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/image-generation-vram-explained/https://localllmchecker.com/guides/image-generation-vram-explained/A diffusion model is a pipeline of parts, and its memory is not the size of one file. Where the gigabytes go in SDXL and Flux, and the 3 settings that decide.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/video-generation-vram-and-render-time/https://localllmchecker.com/guides/video-generation-vram-and-render-time/Video models make one very long sequence, and attention cost explodes with it. Why a 5 second clip is hard, which models are practical, and how long they take.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/local-text-to-speech-hardware/https://localllmchecker.com/guides/local-text-to-speech-hardware/Speech models are small, and what matters is seconds of audio per second of compute. Kokoro, XTTS, F5-TTS and Whisper compared across GPUs.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/rtx-3050-6gb-local-llm-test/https://localllmchecker.com/guides/rtx-3050-6gb-local-llm-test/A real Qwen2.5 7B run on an RTX 3050 6 GB with Ollama: the memory the desktop takes, and the speed cliff when layers spill into RAM. Five findings.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/best-local-llm-for-coding/https://localllmchecker.com/guides/best-local-llm-for-coding/A coding assistant needs a long context window and fast prompt reading more than size. The strongest code model for each amount of VRAM, and what a repo costs.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/how-to-run-a-local-llm/https://localllmchecker.com/guides/how-to-run-a-local-llm/From checking your GPU memory to a working chat in about 15 minutes: install Ollama, pick a model that fits, confirm it uses your GPU, and fix the usual problems.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/guides/run-big-moe-models-with-expert-offload/https://localllmchecker.com/guides/run-big-moe-models-with-expert-offload/A 120B mixture-of-experts model does not need 60 GB of VRAM. Keep its experts in system RAM, find your --n-cpu-moe value, and see what speed to expect.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/We ran Qwen2.5 7B on an RTX 3050 6 GB across five context sizes and six offload levels and compared the results with the calculator. The first version was wrong in three ways: it overestimated memory by 8 to 18 percent, it said a 7B model did not fit a 6 GB card that held it with half a gigabyte to spare, and it mixed binary and decimal gigabytes in the speed formula. All three are fixed. On this machine memory is now off by about 2 percent and speed by about 2.5 percent on average. The CUDA and CPU efficiency constants are now the measured values, 0.79 and 0.63. The full results, the raw data and the script that produced them are on the validation page.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/The KV cache used to assume one attention design per model. It now reads the kind of every layer from the configuration: full, sliding window, linear or compressed. That corrects the cache of hybrid models such as Qwen3.5, which we had overstated by about 4 times, and of models that alternate full and windowed layers, such as gpt-oss and Gemma, which we had understated or overstated depending on the case. The catalog was cut from 133 to 82 models. We removed research baselines, very old and very small models, and every model whose attention layout could not be established from its configuration, since a memory figure we cannot stand behind is worse than none.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/When a mixture-of-experts model does not fit in VRAM, the engine now also considers keeping its experts in system RAM and everything dense on the GPU, which is what llama.cpp does with --cpu-moe, and reports whichever is faster. MoE speeds were too optimistic by about 2 times, because gathering small expert matrices wastes bandwidth that the plain formula counts as used. We now apply a factor of 0.58 to the GPU bandwidth for these models, set from published llama.cpp results. We have not measured a MoE model ourselves and say so wherever those speeds appear.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/gpt-oss-20b and gpt-oss-120b ship their experts in MXFP4, so a download is the same size whatever the file is called. We now use the file size read from the repository, 11.3 GiB and 59 GiB. The 20B model had been shown as fitting a 12 GB card with room to spare, and it does not: its weights alone take nearly all of it.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/The "best model for your card" lists now treat a mixture-of-experts model as sitting between its active and total size, favor newer and more widely downloaded models slightly, and offer only general-purpose models as the all-round pick. Code-only, vision-only and very small models still appear in the full tables. Capability tags that Hugging Face does not provide were filled in by hand for 35 models, and are kept in a curation file so a catalog refresh does not lose them.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/The live calculators on the home, compare and buying pages were sent a trimmed copy of the catalog that left out three fields the engine reads: whether the embedding is tied, and the expert counts of mixture-of-experts models. Their answers could differ from the static page next to them. They now share one packing function, and a test checks that every model gives the same answer before and after packing.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/Qwen2.5 Coder 7B, 14B and 32B and Qwen3 Coder 30B A3B are now listed as models in their own right. The Qwen2.5 and Qwen3 Coder models share an architecture with their general-purpose siblings and need the same memory, but they answer a different question, which is what to run for code, so they have their own pages and appear in the coding picks.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/We added a library of guides on how much VRAM you need, quantization, the KV cache, memory bandwidth, buying hardware, running models in llama.cpp and Ollama, and image, video and speech generation. Each one is built from the same engine as the calculator, so its numbers cannot drift from it.Mon, 05 Oct 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/The media catalog was rebuilt from Hugging Face, with component sizes read from the published files.Wed, 02 Sep 2026 00:00:00 GMThttps://localllmchecker.com/updates/https://localllmchecker.com/updates/The language model catalog was rebuilt from Hugging Face and the price file was reviewed by hand.Tue, 01 Sep 2026 00:00:00 GMT