
Resolution Criteria:
This market resolves YES if Kearm successfully runs DeepSeek-R1, quantized via GGUF but not the R1 based distillations, locally on his home GPU rig by 11:59 PM PT on Feb 9, 2025. Successful execution requires meeting the following criteria:
The GPU rig must be able to answer this prompt via ktransformers without segfaulting or crashing five times in a row to prevent possible variance: "What is the best way for humanity to ascend the Kardashev scale?"
This prompt was chosen as it generates long reasoning chains given it's both vauge yet a decidedly complex problem that is arguably a good fit for a reasoning model.
This main issue's with this problem are outlined in this Github issue: https://github.com/kvcache-ai/ktransformers/issues/117#issuecomment-2637940402
If Kearm fails to achieve this benchmark by the deadline, the market resolves NO, regardless of progress or pending tests.
The specifications of the local GPU rig is here:

Livestream will be running on Twitch primarily at https://www.twitch.tv/kearm
🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ760 | |
| 2 | Ṁ219 | |
| 3 | Ṁ142 | |
| 4 | Ṁ127 | |
| 5 | Ṁ116 |
People are also trading
Attempting the 5 runs in a row live right now at https://www.twitch.tv/kearm and https://x.com/i/broadcasts/1zqJVjQzYyWJB.
Rooting for you in spirit but also keeping in mind what happened last time: https://manifold.markets/Kearm20/will-i-be-able-to-run-deepseekv3-10
@cherrvak Agreed and this one is arguably significantly more difficult and I will be sleeping this time.







