Model Spend Arena2ND ED.
serves endpoints · 2026-08-14

CoreWeave

origin not confirmed

Models it serves

13 endpoints, cheapest first. This is the view the model-by-model tables cannot give you: what one provider actually offers, and at what precision.

ModelQuant$/M in$/M outContextUptime 30m
gpt-oss-20bfp4$0.03$0.13131,07299.6%
Granite 4.1 8Bbf16$0.05$0.10131,072100.0%
gpt-oss-120bfp4$0.03$0.17131,07294.2%
Nemotron 3.5 Lightningbf16$0.10$0.25262,144100.0%
Gemma 4 31Bbf16$0.10$0.34262,14497.0%
DeepSeek V4 Flashfp8$0.13$0.28262,144100.0%
MiniMax-M3fp4$0.23$0.96262,14499.9%
Qwen3.6 35B A3Bfp8$0.25$1.25262,14499.9%
Qwen3.5-35B-A3Bfp8$0.25$1.25262,14488.4%
GLM-5.2fp4$0.76$2.42262,14499.9%
Kimi K2.6fp4$0.65$3.41262,14499.9%
Qwen3.6 27Bfp8$0.60$3.60262,14499.8%
Kimi K2.7 Codeint4$0.71$3.50262,144100.0%
◆ Some links on this page are referral links: if you sign up through them this site may earn a commission, at no extra cost to you. Rankings, prices and measurements are taken from the sources listed under Method, and are not influenced by whether a provider has a referral programme.