40 articles · newest first
A 24 GB card runs Qwen3.8 27B at Q4 or Q5. Q8 needs 32 GB. The surprise is the context cache: 128K tokens costs about 8 GB, a quarter of what an older 32B model charges.
Voltage · 1d ago · 9 min
32GB does not unlock 70B models. It buys better quants and a much longer memory for the 27B to 35B class. Five picks that fit a 5090, with the math shown.
Chiptune · 4d ago · 10 min
Seven mixture-of-experts models, from gpt-oss 20B to Qwen3 235B, split into what stays on your card and what moves to system RAM. The card barely changes. The RAM you need goes from 16 GB to 96 GB.
Voltage · 5d ago · 10 min
You've got a 24GB card and three excellent dense models that fit it. Here's how Mistral Small 24B, Qwen3.6 27B and Gemma 4 31B differ on VRAM, context and modality — and a clear pick for each kind of user.
Tensor · Sep 26 · 7 min
Qwen2.5-Coder-32B still runs a serious private coding assistant on a single 24GB card, with a 32K native context (128K with YaRN). The VRAM math, the tight-but-real fit, and how it stands against the newer Qwen3-Coder.
Tensor · Sep 26 · 7 min
Microsoft's Phi-4 is a compact 14B that runs on a mid-range card — but at 16K context and a late-2024 release, it shows its age next to newer 14B-class models. The VRAM math and an honest verdict.
Tensor · Sep 26 · 7 min
Qwen3.6's dense 27B is the no-fuss 24GB-card option — every parameter fires, nothing to offload, a real 256K context. The VRAM math, the GPU, and how it differs from the 35B-A3B MoE.
Tensor · Sep 26 · 7 min
Qwen3-Next 80B-A3B activates only 3B parameters per token — so it's fast. But 80B of weights still have to live somewhere. The honest VRAM math, the offload trick, and what it really takes.
Tensor · Sep 26 · 7 min
Gemma 4's mixture-of-experts model holds about 26B parameters but activates only 3.8B — so it's quick, and it fits a 16GB card at Q4. The VRAM math, the GPU, and the run command.
Tensor · Sep 26 · 7 min
Gemma 4 31B is the big dense Gemma — the one that asks for a 24GB card and rewards it. The VRAM per quant, the GPU that fits, and how to run it without pulling the wrong tag.
Tensor · Sep 26 · 7 min
Qwen3.6's default is a mixture-of-experts model: 35B on disk, 3B active per token. That makes it fast — but the VRAM math has a twist worth understanding before you buy a card.
Tensor · Sep 26 · 7 min
Type ollama run qwen3.5 and this is the model you get — the 9B. What it is, how little VRAM it needs, and the cheapest card that runs it well.
Tensor · Sep 26 · 7 min
Google's Gemma 4 12B is the multimodal open model that fits a mid-range card — if you pull the right Ollama tag. The honest VRAM math, the cheapest GPU that runs it, and how to start.
Tensor · Sep 26 · 7 min
Unusually modest official specs: a GTX 1660 for 1080p30 and an RTX 3060 Ti for 1440p60. But both assume upscaling, and there's no 4K tier or storage size yet.
Hitbox · Sep 25 · 5 min
The official specs have a hard floor: no hardware ray tracing, no game. Which cards clear it, why GTX 10 and 16 series don't, and what each tier actually asks for.
Hitbox · Sep 25 · 6 min
Ollama's default gemma4 is the small E4B and qwen3.5 is the 9B, not what most people assume. What each tag really downloads, what the bigger models need at the 64K context agents use, and what to pick for 12, 16 and 24GB cards.
Tensor · Sep 25 · 7 min
Everyone says 16GB minimum. That's only true if the whole model has to live on the GPU. llama.cpp's own guide runs it on an 8GB RTX 2060. Here are the commands, and the starting values for your card.
Tensor · Sep 25 · 7 min
GLM-4.5-Air is 106B total but only 12B active, ~60GB at 4-bit — small enough for one 80GB card or a 128GB unified box. Exactly what hardware runs it, and the cheapest way if you don't own one.
Tensor · Sep 24 · 7 min
GLM ties Kimi K3 at the top of the open-weight rankings, but 'run it locally' depends on which GLM — from the 106B Air that fits one card to the 744B GLM-5 that needs a 256GB Mac or the cloud. The honest map, with real numbers.
Tensor · Sep 24 · 8 min
Mistral Small 24B fits a single 24GB card at Q4 (~14GB). VRAM by quant, which GPU, realistic speeds by runtime, and the cheapest way to run it — own or rent.
Tensor · Sep 24 · 7 min
For local AI the decider is VRAM — and that's exactly where laptops are capped. Why a desktop (or a unified-memory mini-PC) usually wins, when a gaming laptop is still right, and how to choose.
Voltage · Sep 22 · 8 min
Quantization is the only reason Scout runs at home — but MoE means you shrink all 109B, not the 17B active. The honest GGUF memory ladder, the Q4_K_M sweet spot, and how low you can safely go.
Tensor · Sep 22 · 8 min
Scout is mixture-of-experts, so you're buying for 109B in memory, not 17B active. The honest GPU ladder — why a 32GB card is the sweet spot, when 24GB works, and when unified memory beats them all.
Voltage · Sep 22 · 8 min
Quantization is the trick that lets a 16GB card run models it has no business running. Here's how GGUF quant levels work, which one I default to, the memory math, and the honest quality trade — so you get the most model your VRAM can hold.
Tensor · Sep 22 · 9 min
12GB of VRAM at ~$249, and the one card the 2026 price surge barely touched. Superb budget hardware for local LLMs — if you can live with Intel's software tax. My honest take on who should buy one.
Voltage · Sep 22 · 8 min
OpenAI shipped two very different models. The 20B fits a normal 16GB card and flies; the 120B is an 80GB-class model you'll mostly rent. Here's the card I'd buy for each — VRAM-first, with current requirements cited.
Tensor · Sep 22 · 8 min
Two used RTX 3090s give you 48GB of VRAM for a third of a 5090's price — so is a dual-GPU build the smart move for local AI? The honest math, whether models really split across cards, and the power-and-complexity costs the sticker price hides.
Tensor · Sep 22 · 8 min
Qwen3 isn't one model — it's a family from 4B to 235B, and the right GPU depends entirely on which you run. My VRAM-first map of every variant, the MoE trap that trips people up, and the used card I'd hand most people.
Tensor · Sep 22 · 8 min
At ~$4,600 for 32GB of VRAM, the RTX 5090 is either the only card that makes sense — or a tax on headroom you'll never use. An honest, VRAM-first buy call with the cheaper alternatives priced in.
Tensor · Sep 22 · 8 min
Same $549 launch price, genuinely different cards — and 2026's price surge pulled them apart. The RX 9070 wins raster and VRAM; the RTX 5070 wins ray tracing and DLSS 4. Here's how I'd choose for gaming and local AI.
Voltage · Sep 22 · 7 min
I priced out every rung of the local-AI hardware ladder this week. One tier is quietly the best deal on the list, and one isn't worth it at any price.
Voltage · Aug 27 · 9 min
$501 off and two free games sounds great until you hit the line that says 16GB of RAM — and fixing that yourself right now costs almost as much as the discount.
Ledger · Aug 27 · 7 min
A handful of llama.cpp flags let a 16GB card run models that shouldn't fit, and most people running local AI have never touched them.
Tensor · Aug 25 · 7 min
Modders in China are selling RTX 3080s with double the stock VRAM for about $445. Here's the real math on whether that's a shortcut or a way to fry your PC.
Voltage · Aug 24 · 7 min
The RTX 5050 is the cheap pick and the RX 9070 is the overall pick. Only one of them still makes sense a year from now.
Voltage · Aug 24 · 7 min
A new hardware breakdown puts real numbers on the two components most local-AI buying guides skip entirely — and skimping on either turns a snappy agent into a 10-minute wait.
Tensor · Aug 19 · 7 min
the 16GB version is up 88% since launch. the 8GB version barely moved. here's the real math if you're buying for local AI, not just gaming.
Voltage · Aug 16 · 7 min
The one rule is buy VRAM, not benchmarks — and 2026's price surge is now cooling. My current, honest map: which card for which budget (Arc B580 to RTX 5090), what each actually runs, and when to rent or go unified-memory instead.
Tensor · Aug 14 · 11 min
The RTX PRO 6000 nearly doubled in eight months. Here's what your money actually buys against every real alternative for running big models at home.
Voltage · Aug 14 · 8 min
The card was reportedly finished and sitting with board partners — and Nvidia shelved the whole Super lineup anyway.
Voltage · Aug 9 · 7 min