buyaigear
Prices checked 8 min ago

What runs each model, and what it costs today

28 reference models, each matched against every GPU and desktop box we track. The cheapest machine is the lowest live price, item plus shipping, that fits the model at Q4_K_M with 8K of context. Open a model for every quantization and every machine.

ModelReleasedNeeds at Q4_K_MCheapest machine that fits todayPrice
DeepSeek
DeepSeek V4.1 Flash 552B-A16B 552B total, 16B activeSep 2026338 GBApple Mac Studio M3 Ultra, 512 GB$18,566.48
DeepSeek V4 Flash 284B-A13B 284B total, 13B activeJul 2026175 GBApple Mac Studio M3 Ultra, 256 GB$8,499
DeepSeek V3.1 671B-A37B 671B total, 37B activeAug 2025411 GBApple Mac Studio M3 Ultra, 512 GB$18,566.48
Gemma
Gemma 4 12B 11.9BMay 20269 GBNVIDIA Tesla P100 16GB$127.63
Gemma 4 31B 30.7BMar 202621 GBNVIDIA Tesla P40$273
Gemma 4 26B-A4B 25.2B total, 3.8B activeMar 202617 GBNVIDIA Tesla P40$273
Gemma 3 27B 27.4BMar 202519 GBNVIDIA Tesla P40$273
GLM
GLM-5.3 753B-A40B 753B total, 40B activeAug 2026462 GBApple Mac Studio M3 Ultra, 512 GB$18,566.48
GLM-5.3 Flash 320B-A18B 320B total, 18B activeAug 2026197 GBApple Mac Studio M3 Ultra, 256 GB$8,499
GLM-4.6 355B-A32B 355B total, 32B activeSep 2025221 GBApple Mac Studio M3 Ultra, 256 GB$8,499
GLM-4.5 Air 106B-A12B 106B total, 12B activeJul 202568 GBFramework Desktop, Ryzen AI Max+ 395, 128 GB$3,464.65
gpt-oss
gpt-oss-120b 116.8B total, 5.1B activeAug 202573 GBFramework Desktop, Ryzen AI Max+ 395, 128 GB$3,464.65
gpt-oss-20b 21B total, 3.6B activeAug 202515 GBNVIDIA Tesla P100 16GB$127.63
Llama
Llama 4 Scout 109B-A17B 109B total, 17B activeApr 202570 GBFramework Desktop, Ryzen AI Max+ 395, 128 GB$3,464.65
Llama 3.3 70B 70.6BDec 202447 GBFramework Desktop, Ryzen AI Max+ 395, 128 GB$3,464.65
Llama 3.1 8B 8BJul 20247 GBNVIDIA Tesla P100 16GB$127.63
MiMo
MiMo V2.6 Pro 1.02T-A42B 1020B total, 42B activeSep 2026624 GBnothing we track fits
MiMo V2.6 Flash 309B-A15B 309B total, 15B activeSep 2026190 GBApple Mac Studio M3 Ultra, 256 GB$8,499
Mistral
Mistral Medium 3.5 128B 128BMar 202683 GBFramework Desktop, Ryzen AI Max+ 395, 128 GB$3,464.65
Mistral Small 4 119B-A6.5B 119B total, 6.5B activeMar 202679 GBFramework Desktop, Ryzen AI Max+ 395, 128 GB$3,464.65
Mistral Small 3.2 24B 24BJun 202517 GBNVIDIA Tesla P40$273
Qwen
Qwen3.8 Flash-Next 125B-A6B 125B total, 6B activeAug 202678 GBFramework Desktop, Ryzen AI Max+ 395, 128 GB$3,464.65
Qwen3.8 27B 27BAug 202619 GBNVIDIA Tesla P40$273
Qwen3.6 27B 27BApr 202619 GBNVIDIA Tesla P40$273
Qwen3.6 35B-A3B 35B total, 3B activeApr 202623 GBNVIDIA Tesla P40$273
Qwen3 Coder 480B-A35B 480B total, 35B activeJul 2025296 GBApple Mac Studio M3 Ultra, 512 GB$18,566.48
Qwen3 235B-A22B 235B total, 22B activeJul 2025146 GBApple Mac Studio M3 Ultra, 256 GB$8,499
Qwen3 32B 32.8BApr 202524 GBNVIDIA Tesla P40$273

Needs includes the KV cache for 8K of context and runtime overhead. How we estimate.