I have forsale two Nvidia Tesla t4 16gb gpu. These are strictly for compute, and are used for machine learning workloads. You can use these with 14 to 20 billion parameters LLM models (if you use both in ollama you can get to 40b parameter models). Both are working excellent. I also have a small low profile bracket (but only one). Both are with full size brackets now. You can use Gemma3, qwen, DeepSeek, and other LLMs with one or both of these.
These are passively cooled.
$500 each, $900 for both. Cash, local pickup only.