What can your hardware run?
Pick your GPU. Get every open-weight model that fits at its best quantization — and the cheapest verified host for everything that doesn't.
scraped nightlyno estimatesfull price history
Not running locally? See who hosts it cheapest →Your rig24 GB VRAM
1of 5 models fit your RTX 4090
at their best quantization
at their best quantization
Llama 3.1 8B Instruct
params × bytes/quant × 1.2 overhead · verified May 27, 2026
The market today
all figures verified May 27, 2026Price movers · 24hChangelog →
Llama 3.1 8B Instruct$/1M out
— new low — new lowDeepSeek V3$/1M out
— new low — new lowQwen 2.5 72B Instruct$/1M out
— new low — new lowLlama 3.1 70B Instruct$/1M out
— new low — new lowFeatured · 90-day price$/1M output
Llama 3.1 8B Instruct$0.0200 · ▼ 60% / 90d
cheapest verified host per day · source: nightly scrape · full history →
Every provider, tracked nightly
All 7 providers →Latest changes
Full changelog + RSS →May 27DeepInfra listed DeepSeek V3 at $0.32/1M out
May 27DeepInfra listed Llama 3.1 70B Instruct at $0.40/1M out
May 27DeepInfra listed DeepSeek R1 Distill Llama 70B at $0.70/1M out
May 27DeepInfra listed Llama 3.1 8B Instruct at $0.0200/1M out
May 27DeepInfra listed Qwen 2.5 72B Instruct at $0.36/1M out
Price alerts
One email when a model you track gets cheaper. No digest, no spam.
confirm by email · unsubscribe any time