cdc
CDC
5 results
VM Types
t4-large
Engines
ollama
GPUs
NVIDIA Tesla T4
Compute
ENTRY GPU + CPU
Entry GPU (16 GB class)
cross-foundation ranking →single entry datacenter card — Tesla T4 / L4 class: the 'runs on what you already have' tier; ranked by usability because item budgets bite here
#
Model
Hardware – Engine
Usability
Completed
tok/s
Capability
Date
1
hf.co/unsloth/gpt-oss-20b-GGUF:UD-Q4_K_XL
OpenAIq4
1×NVIDIA Tesla T4 – ollama
cdc
0.76±0.072
0.81 capability
97%
160/165 · 5 timeouts
37.3
Aug 31
2
hf.co/google/gemma-4-12B-it-qat-q4_0-gguf:latest⚠
Googleq4
1×NVIDIA Tesla T4 – ollama
cdc
0.53±0.129
0.66 capability
60%
99/165 · 66 timeouts · 27 not attempted (6 h run cap)
17.3
Aug 30
3
hf.co/google/gemma-4-12B-it-qat-q4_0-gguf:latest⚠
Googleq4
1×NVIDIA Tesla T4 – ollama
cdc
0.50±0.149
0.64 capability
63%
104/165 · 61 timeouts · 22 not attempted (6 h run cap)
17.4
Aug 30
no accelerator — Xeon-class CPU inference; ranked by usability because latency is part of the result
#
Model
Hardware – Engine
Usability
Completed
tok/s
Capability
Date
1
hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q4_K_M⚠
cpu-4xlarge – ollama
cdc
0.67±0.110
0.74 capability
82%
124/152 · 28 timeouts · 10 not attempted (6 h run cap)
14.3
Aug 31
2
hf.co/google/gemma-4-26B-A4B-it-qat-q4_0-gguf:latest⚠
Googleq4
cpu-4xlarge – ollama
cdc
0.24±0.156
0.27 capability
25%
42/165 · 123 timeouts · 59 not attempted (6 h run cap)
12.6
Aug 30