GPU Sizing Studio · Model directorySign in

Model directory

29 models with published architecture specs. No account needed — pick one to see full specs, or sign in to size a deployment for it.

DeepSeek

DeepSeek-V3.1
37B active / 671B total · 8/256 experts
163,840 token context · FP8 native
DeepSeek-V3
37B active / 671B total · 8/256 experts
131,072 token context · FP8 native
DeepSeek-R1
37B active / 671B total · 8/256 experts
131,072 token context · FP8 native

GPT-OSS

gpt-oss-120b
5.1B active / 117B total · 4/128 experts
131,072 token context · BF16 native
gpt-oss-20b
3.6B active / 21B total · 4/32 experts
131,072 token context · BF16 native

Gemma

Gemma 3 27B
27B params
131,072 token context · BF16 native
Gemma 2 9B
9.24B params
8,192 token context · BF16 native
Gemma 2 27B
27.2B params
8,192 token context · BF16 native

Llama

Llama 4 Scout (17B active / 109B total)
17B active / 109B total · 1/16 experts
10,485,760 token context · BF16 native
Llama 4 Maverick (17B active / 400B total)
17B active / 400B total · 1/128 experts
1,048,576 token context · BF16 native
Llama 3.1 8B
8.03B params
131,072 token context · BF16 native
Llama 3.1 70B
70.6B params
131,072 token context · BF16 native
Llama 3.3 70B
70.6B params
131,072 token context · BF16 native
Llama 3.1 405B
405B params
131,072 token context · BF16 native

Mistral

Mistral Large (2411)
123B params
131,072 token context · BF16 native
Mistral 7B v0.3
7.25B params
32,768 token context · BF16 native
Mixtral 8x7B
12.9B active / 46.7B total · 2/8 experts
32,768 token context · BF16 native
Mixtral 8x22B
39B active / 141B total · 2/8 experts
65,536 token context · BF16 native

Phi

Phi-3 Mini 3.8B
3.82B params
131,072 token context · BF16 native
Phi-3 Medium 14B
14B params
131,072 token context · BF16 native

Qwen

Qwen3 32B
32.8B params
40,960 token context · BF16 native
Qwen3 30B-A3B
3.3B active / 30.5B total · 8/128 experts
40,960 token context · BF16 native
Qwen3 235B-A22B
22B active / 235B total · 8/128 experts
40,960 token context · BF16 native
Qwen3.6 27B
27B params
262,144 token context · BF16 native
Qwen3.6 35B-A3B
3B active / 35B total · 9/256 experts
262,144 token context · BF16 native
Qwen2.5 7B
7.6B params
131,072 token context · BF16 native
Qwen2.5 14B
14.7B params
131,072 token context · BF16 native
Qwen2.5 32B
32.5B params
131,072 token context · BF16 native
Qwen2.5 72B
72.7B params
131,072 token context · BF16 native