Observed arrival · 2026-10-10
GPUCostLab shows its LLM cost arithmetic
A set of free calculators and reference tables for estimating LLM memory, GPU costs, fine-tuning, and self-hosting versus API use.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People estimating LLM hosting and GPU costs
- Worth noticing
- The fine-tuning estimator exposes throughput as an editable assumption, while the VRAM tool accounts for several attention architectures.
Field notes
The VRAM calculation combines model weights with a KV cache that grows with context and batch, using each model’s config.json; the page names GQA, MLA, sliding-window, and linear-attention handling. Its GPU table lists on-demand GPU-hour prices across eleven GPU clouds alongside AWS, Google Cloud, and Azure, with read dates and source links. Fine-tuning throughput is presented as an editable assumption rather than a measurement.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue