LLM Hardware Checker

Free tool. Detect your Mac or GPU and see which of 27 open-weight LLMs (Llama, Qwen, Gemma, Mistral, DeepSeek) you can run smoothly — with realistic tokens/sec estimates from 80+ measured benchmarks.

Browser-based. No signup. 82+ measured benchmarks across Apple Silicon, NVIDIA, and AMD hardware → an honest verdict on which open-weight models will actually run smoothly on your machine, with realistic tokens/sec estimates.

Detecting your hardware…

What the LLM hardware checker does

Answer "can my machine run this model?" without guessing. Detect your Mac or GPU (or pick one), and this tool shows which of its 27 open-weight models released 2024–2025, from SmolLM2 1.7B to DeepSeek-V3 671B (MoE) you can run smoothly, with realistic tokens/sec drawn from 80+ measured benchmarks rather than spec-sheet estimates.

How to check your hardware

  1. Let it detect your Mac/GPU, or select your setup.
  2. See which models fit and how fast they'll actually run.
  3. Use the tokens/sec estimate to judge whether it's fast enough for your use.

What decides whether a model runs

Two limits. First, memory: the model's weights (at your quantization) must fit in VRAM or unified memory, or it won't load. Second, memory bandwidth: once it fits, bandwidth sets how fast tokens come out. That's why an Apple M-series with lots of unified memory can run larger models than a GPU with less VRAM. To go the other way — pick hardware for a target model — use the hardware recommender; to decide whether to buy at all, the self-host vs API calculator; raw numbers live in the benchmark database.

Frequently asked questions

How much RAM or VRAM do I need?

Roughly the parameter count × bytes-per-parameter at your quantization — an 8B model at 4-bit needs ~5–6 GB; the checker tells you exactly what fits.

Why does Apple Silicon punch above its weight?

Its unified memory is large and shared with the GPU, so it can hold bigger models than a discrete GPU with less VRAM — even at similar raw speed.

Where do the speeds come from?

80+ real measured benchmarks, not vendor claims.

Related guides

SharePost

More tools like this

LLM API Price Tracker

AI & LLM

LLM API pricing per 1M tokens across Anthropic, OpenAI, Google, Meta, Mistral, DeepSeek, xAI. Sortable, source-checked, with daily snapshots.

Open

Embedding Model Prices

AI & LLM

Embedding API cost per 1M tokens, vector dimensions and max input for OpenAI, Voyage, Cohere and Gemini models.

Open

GPU / Mac Recommender

AI & LLM

Pick a model + your minimum tokens/sec. Get the cheapest Mac M-series or NVIDIA / AMD GPU that runs it smoothly.

Open
Take a break Play 2048 Evolved From the blog AI systems, deep-dived