M3 Ultra 96GB

77 models / sorted by best_quality

Quality Fastest Least mem Newest
Runs comfortably
Qwen3.8-Flash-Next UD-IQ1_S Unsloth
72.5GB 262k ctx
FIT Runs comfortably
SPEED 141 tok/s
ⓘ estimated
MEMORY 73/96GB
Ornith-1.5-35B-A3B BF16
71.1GB 262k ctx
FIT Runs comfortably
SPEED 83 tok/s
ⓘ estimated
MEMORY 71/96GB
Ornith-1.5-35B-A3B Q8_0
37.8GB 262k ctx
FIT Runs comfortably
SPEED 156 tok/s
ⓘ estimated
MEMORY 38/96GB
Ornith-1.5-35B-A3B Q6_K
29.2GB 262k ctx
FIT Runs comfortably
SPEED 202 tok/s
ⓘ estimated
MEMORY 29/96GB
Ornith-1.5-35B-A3B Q5_K_M
25.3GB 262k ctx
FIT Runs comfortably
SPEED 233 tok/s
ⓘ estimated
MEMORY 25/96GB
Ornith-1.5-35B-A3B Q4_K_M
21.7GB 262k ctx
FIT Runs comfortably
SPEED 271 tok/s
ⓘ estimated
MEMORY 22/96GB
Qwen3.6 35B A3B Q4_K_M
20.0GB 262k ctx
FIT Runs comfortably
SPEED 286 tok/s
ⓘ estimated
MEMORY 20/96GB
Qwen3.6 35B A3B Q8_0
37.0GB 262k ctx
FIT Runs comfortably
SPEED 155 tok/s
ⓘ estimated
MEMORY 37/96GB
Pokee-Isaac 28B Q4_K_M
16.0GB 10000k ctx
FIT Runs comfortably
SPEED 31 tok/s
ⓘ estimated
MEMORY 16/96GB
Pokee-Isaac 28B Q8_0
28.0GB 10000k ctx
FIT Runs comfortably
SPEED 18 tok/s
ⓘ estimated
MEMORY 28/96GB
Pokee-Isaac 28B FP16
56.0GB 10000k ctx
FIT Runs comfortably
SPEED 9 tok/s
ⓘ estimated
MEMORY 56/96GB
Qwen3 235B A22B Q2_K
68.0GB 128k ctx
FIT Runs comfortably
SPEED 77 tok/s
ⓘ estimated
MEMORY 68/96GB
Qwen3.8-27B NVFP4 Unsloth
15.2GB 262k ctx
FIT Runs comfortably
SPEED 32 tok/s
ⓘ estimated
MEMORY 15/96GB
Qwen3.8-27B UD-IQ1_S Unsloth
6.2GB 262k ctx
FIT Runs comfortably
SPEED 79 tok/s
ⓘ estimated
MEMORY 6/96GB
Qwen3.8-27B UD-Q2_K_XL Unsloth
9.8GB 262k ctx
FIT Runs comfortably
SPEED 50 tok/s
ⓘ estimated
MEMORY 10/96GB
Qwen3.8-27B UD-Q4_K_XL Unsloth
17.6GB 262k ctx
FIT Runs comfortably
SPEED 28 tok/s
ⓘ estimated
MEMORY 18/96GB
Gemma 4 31B Q4_K_M
18.3GB 256k ctx
FIT Runs comfortably
SPEED 27 tok/s
ⓘ estimated
MEMORY 18/96GB
Gemma 4 31B NVFP4
16.0GB 256k ctx
FIT Runs comfortably
SPEED 31 tok/s
ⓘ estimated
MEMORY 16/96GB
Gemma 4 31B BF16
61.4GB 256k ctx
FIT Runs comfortably
SPEED 8 tok/s
ⓘ estimated
MEMORY 61/96GB
Gemma 4 31B QAT
18.0GB 256k ctx
FIT Runs comfortably
SPEED 27 tok/s
ⓘ estimated
MEMORY 18/96GB
Gemma 4 31B Q8_0
32.6GB 256k ctx
FIT Runs comfortably
SPEED 15 tok/s
ⓘ estimated
MEMORY 33/96GB
Gemma 4 26B A4B NVFP4
14.0GB 256k ctx
FIT Runs comfortably
SPEED 232 tok/s
ⓘ estimated
MEMORY 14/96GB
Gemma 4 26B A4B Q4_K_M
13.0GB 256k ctx
FIT Runs comfortably
SPEED 250 tok/s
ⓘ estimated
MEMORY 13/96GB
Gemma 4 26B A4B Q8_0
25.0GB 256k ctx
FIT Runs comfortably
SPEED 130 tok/s
ⓘ estimated
MEMORY 25/96GB
Gemma 4 26B A4B BF16
50.0GB 256k ctx
FIT Runs comfortably
SPEED 65 tok/s
ⓘ estimated
MEMORY 50/96GB
Qwen3.6 27B Q8_0
28.0GB 128k ctx
FIT Runs comfortably
SPEED 18 tok/s
ⓘ estimated
MEMORY 28/96GB
Qwen3.6 27B Q4_K_M
16.5GB 128k ctx
FIT Runs comfortably
SPEED 30 tok/s
ⓘ estimated
MEMORY 17/96GB
Qwen3 32B Q8_0
34.0GB 128k ctx
FIT Runs comfortably
SPEED 14 tok/s
ⓘ estimated
MEMORY 34/96GB
Qwen3 32B Q4_K_M
20.0GB 128k ctx
FIT Runs comfortably
SPEED 25 tok/s
ⓘ estimated
MEMORY 20/96GB
Qwen3 32B Q6_K
28.5GB 128k ctx
FIT Runs comfortably
SPEED 17 tok/s
ⓘ estimated
MEMORY 29/96GB
Qwen3 32B Q5_K_M
24.5GB 128k ctx
FIT Runs comfortably
SPEED 20 tok/s
ⓘ estimated
MEMORY 25/96GB
Ornith-1.5-9B BF16
18.4GB 262k ctx
FIT Runs comfortably
SPEED 27 tok/s
ⓘ estimated
MEMORY 18/96GB
Ornith-1.5-9B Q8_0
9.8GB 262k ctx
FIT Runs comfortably
SPEED 50 tok/s
ⓘ estimated
MEMORY 10/96GB
Ornith-1.5-9B Q6_K
7.6GB 262k ctx
FIT Runs comfortably
SPEED 65 tok/s
ⓘ estimated
MEMORY 8/96GB
Nemotron 3.5 Lightning FP16
60.0GB 1000k ctx
FIT Runs comfortably
SPEED 72 tok/s
ⓘ estimated
MEMORY 60/96GB
Nemotron 3.5 Lightning Q8_0
58.0GB 1000k ctx
FIT Runs comfortably
SPEED 74 tok/s
ⓘ estimated
MEMORY 58/96GB
Nemotron 3.5 Lightning Q4_K_M
32.0GB 1000k ctx
FIT Runs comfortably
SPEED 135 tok/s
ⓘ estimated
MEMORY 32/96GB
Ornith-1.5-9B Q5_K_M
6.6GB 262k ctx
FIT Runs comfortably
SPEED 74 tok/s
ⓘ estimated
MEMORY 7/96GB
Ornith-1.5-9B Q4_K_M
5.8GB 262k ctx
FIT Runs comfortably
SPEED 85 tok/s
ⓘ estimated
MEMORY 6/96GB
Qwen3 30B A3B Q8_0
31.0GB 128k ctx
FIT Runs comfortably
SPEED 161 tok/s
ⓘ estimated
MEMORY 31/96GB
Qwen3 30B A3B Q4_K_M
15.5GB 128k ctx
FIT Runs comfortably
SPEED 322 tok/s
ⓘ estimated
MEMORY 16/96GB
Phi-4 14B FP16
29.0GB 16k ctx
FIT Runs comfortably
SPEED 17 tok/s
ⓘ estimated
MEMORY 29/96GB
Phi-4 14B Q8_0
15.0GB 16k ctx
FIT Runs comfortably
SPEED 33 tok/s
ⓘ estimated
MEMORY 15/96GB
Phi-4 14B Q4_K_M
8.5GB 16k ctx
FIT Runs comfortably
SPEED 58 tok/s
ⓘ estimated
MEMORY 9/96GB
Xing4.0-29B-A4B Q4_K_M
20.1GB 256k ctx
FIT Runs comfortably
SPEED 177 tok/s
ⓘ estimated
MEMORY 20/96GB
Xing4.0-29B-A4B BF16
62.4GB 256k ctx
FIT Runs comfortably
SPEED 57 tok/s
ⓘ estimated
MEMORY 62/96GB
Llama 3.3 70B Instruct Q4_K_M
42.0GB 128k ctx
FIT Runs comfortably
SPEED 12 tok/s
ⓘ estimated
MEMORY 42/96GB
Llama 3.3 70B Instruct Q3_K_M
30.0GB 128k ctx
FIT Runs comfortably
SPEED 16 tok/s
ⓘ estimated
MEMORY 30/96GB
Llama 3.3 70B Instruct Q2_K
21.5GB 128k ctx
FIT Runs comfortably
SPEED 23 tok/s
ⓘ estimated
MEMORY 22/96GB
Llama 3.3 70B Instruct Q8_0
74.0GB 128k ctx
FIT Runs comfortably
SPEED 7 tok/s
ⓘ estimated
MEMORY 74/96GB
Qwen3 14B Q8_0
15.0GB 128k ctx
FIT Runs comfortably
SPEED 33 tok/s
ⓘ estimated
MEMORY 15/96GB
Qwen3 14B Q4_K_M
9.0GB 128k ctx
FIT Runs comfortably
SPEED 55 tok/s
ⓘ estimated
MEMORY 9/96GB
Gemma 4 12B Q8_0
13.0GB 128k ctx
FIT Runs comfortably
SPEED 38 tok/s
ⓘ estimated
MEMORY 13/96GB
Gemma 4 12B Q4_K_M
7.2GB 128k ctx
FIT Runs comfortably
SPEED 68 tok/s
ⓘ estimated
MEMORY 7/96GB
Gemma 3 27B Q8_0
28.0GB 128k ctx
FIT Runs comfortably
SPEED 18 tok/s
ⓘ estimated
MEMORY 28/96GB
Gemma 3 27B Q4_K_M
17.0GB 128k ctx
FIT Runs comfortably
SPEED 29 tok/s
ⓘ estimated
MEMORY 17/96GB
Qwen3 8B Q8_0
8.5GB 128k ctx
FIT Runs comfortably
SPEED 58 tok/s
ⓘ estimated
MEMORY 9/96GB
Qwen3 8B Q4_K_M
5.2GB 128k ctx
FIT Runs comfortably
SPEED 94 tok/s
ⓘ estimated
MEMORY 5/96GB
Mixtral 8x7B Instruct Q3_K_M
19.0GB 32k ctx
FIT Runs comfortably
SPEED 94 tok/s
ⓘ estimated
MEMORY 19/96GB
Mixtral 8x7B Instruct Q4_K_M
26.5GB 32k ctx
FIT Runs comfortably
SPEED 67 tok/s
ⓘ estimated
MEMORY 27/96GB
Gemma 3 12B Q8_0
13.5GB 128k ctx
FIT Runs comfortably
SPEED 36 tok/s
ⓘ estimated
MEMORY 14/96GB
Gemma 3 12B Q4_K_M
7.5GB 128k ctx
FIT Runs comfortably
SPEED 66 tok/s
ⓘ estimated
MEMORY 8/96GB
Mistral Nemo 12B Q8_0
13.2GB 128k ctx
FIT Runs comfortably
SPEED 37 tok/s
ⓘ estimated
MEMORY 13/96GB
Mistral Nemo 12B Q4_K_M
7.3GB 128k ctx
FIT Runs comfortably
SPEED 67 tok/s
ⓘ estimated
MEMORY 7/96GB
Qwen3 4B Q4_K_M
2.6GB 32k ctx
FIT Runs comfortably
SPEED 189 tok/s
ⓘ estimated
MEMORY 3/96GB
Phi-4 Mini Q8_0
4.1GB 128k ctx
FIT Runs comfortably
SPEED 120 tok/s
ⓘ estimated
MEMORY 4/96GB
Phi-4 Mini Q4_K_M
2.5GB 128k ctx
FIT Runs comfortably
SPEED 197 tok/s
ⓘ estimated
MEMORY 3/96GB
Llama 3.1 8B Instruct Q4_K_M
4.5GB 128k ctx
FIT Runs comfortably
SPEED 109 tok/s
ⓘ estimated
MEMORY 5/96GB
Llama 3.1 8B Instruct Q8_0
8.0GB 128k ctx
FIT Runs comfortably
SPEED 61 tok/s
ⓘ estimated
MEMORY 8/96GB
Llama 3.1 8B Instruct FP16
16.0GB 128k ctx
FIT Runs comfortably
SPEED 31 tok/s
ⓘ estimated
MEMORY 16/96GB
Gemma 3 4B Q4_K_M
3.0GB 128k ctx
FIT Runs comfortably
SPEED 164 tok/s
ⓘ estimated
MEMORY 3/96GB
Gemma 3 4B Q8_0
5.0GB 128k ctx
FIT Runs comfortably
SPEED 98 tok/s
ⓘ estimated
MEMORY 5/96GB
Llama 3.2 3B Instruct Q8_0
3.5GB 128k ctx
FIT Runs comfortably
SPEED 140 tok/s
ⓘ estimated
MEMORY 4/96GB
Llama 3.2 3B Instruct Q4_K_M
2.0GB 128k ctx
FIT Runs comfortably
SPEED 246 tok/s
ⓘ estimated
MEMORY 2/96GB
Muse Glimmer 30B Q4_K_M
18.0GB 128k ctx
FIT Runs comfortably
SPEED 27 tok/s
ⓘ estimated
MEMORY 18/96GB
Muse Glimmer 30B Q8_0
33.0GB 128k ctx
FIT Runs comfortably
SPEED 15 tok/s
ⓘ estimated
MEMORY 33/96GB
Runs tight
Muse Glimmer 30B FP16
60.0GB 0k cap
FIT Runs tight
SPEED 8 tok/s
ⓘ estimated
MEMORY 60/96GB

Coding alternatives

Hosted coding assistants for comparison - per seat

Cursor Hobby
AI-first code editor. Uses frontier models (GPT-4, Claude). $20/user Business...
$0.0
/MONTH
GitHub Copilot Individual
AI coding assistant by GitHub. GPT-4o and Claude-based completions. $19/seat ...
$10.0
/MONTH
GitHub Copilot Business
AI coding assistant by GitHub. GPT-4o and Claude-based completions. $19/seat ...
$19.0
/MONTH
Cursor Pro
AI-first code editor. Uses frontier models (GPT-4, Claude). $20/user Business...
$20.0
/MONTH
GitHub Copilot Enterprise
AI coding assistant by GitHub. GPT-4o and Claude-based completions. $19/seat ...
$39.0
/MONTH
Cursor Business
AI-first code editor. Uses frontier models (GPT-4, Claude). $20/user Business...
$40.0
/MONTH

Per-token monthly figures are estimates at typical usage. Prices verified periodically - always confirm current pricing on the provider's website.

On a monthly budget? Compare owning hardware vs renting cloud →

Sign in with GitHub to save this rig and get weekly model updates

Sign in with GitHub