GPU Tools
Free, live-priced tools for sizing GPU workloads.
Training Cost Calculator
Estimate how long and how much it costs to pretrain a language model, priced against live GPU rates. Shows the full FLOPs formula.
GPU Recommender
Pick any popular Hugging Face model and get its VRAM requirement at native, FP8, and INT4 precision, plus the cheapest live GPU that fits.
Popular models
Qwen/Qwen3-0.6B
0.8B params · BF16
Qwen/Qwen3-8B
8.2B params · BF16
Qwen/Qwen2.5-1.5B-Instruct
1.5B params · BF16
openai-community/gpt2
0.1B params · F32
Qwen/Qwen2.5-7B-Instruct
7.6B params · BF16
meta-llama/Llama-3.2-1B-Instruct
1.2B params · BF16
Qwen/Qwen3-Embedding-0.6B
0.6B params · BF16
deepseek-ai/DeepSeek-R1
684.5B params · F8_E4M3
Qwen/Qwen3-32B
32.8B params · BF16
meta-llama/Llama-3.1-8B-Instruct
8.0B params · BF16
Qwen/Qwen3-1.7B
2.0B params · BF16
Qwen/Qwen2.5-3B-Instruct
3.1B params · BF16
