⬢github · 160 ★ +49 since we first saw it · pushed 1 h ago · CC-BY-4.0
Magnvs7/Promethevs
PROMETHEVS: run the full DeepSeek-V4.1-Flash-1017 (510 GB) on a gaming PC with 64 GB RAM and a 12 GB GPU. PROMETHEVS engine (compress-then-correct), OpenAI-compatible local server, report and logs.
PROMETHEVS is an open-source engine that runs the full 510 GB DeepSeek-V4.1-Flash model on a modest gaming PC with 64 GB RAM and a 12 GB GPU. Its compress-then-correct method computes tokens on a 2-bit expert copy and recomputes uncertain ones with the original weights, serving results through an OpenAI-compatible local server.
Why now: It was posted on Hacker News for claiming to run an unrunnable-size frontier model locally on consumer hardware, with audited accuracy and speed numbers backing the claim.
Who it is for: Hobbyists and developers who want to run a very large open LLM locally on consumer hardware without trusting cloud providers.
consumer-hardwaredeepseekdeepseek-v4llmllm-inferencelocal-inferencelocal-llmmixture-of-expertsopenai-compatiblequantization
Stars over our 286 snapshots: 111 to 160, since 2 d ago.
Where people talked about it
API: https://socialmediatrends-api.osmike.com/v1/repos/Magnvs7/Promethevs