AI platform teams
LLM VRAM calculator
How much GPU memory a model needs for its weights, key-value (KV) cache and runtime, and which GPUs fit it, with tensor-parallel sizes.
- Presets for Qwen3.8, GLM-5.3, DeepSeek, gpt-oss, Llama, Gemma and more
- Weights, KV cache and recurrent state priced from each model's config.json
- 12 NVIDIA GPUs from L4 to B200, plus Macs up to 512 GB