an LLM as capable as GPT-4 will run on a 3090 before March 2024
23
Ṁ430Ṁ4.8kresolved Mar 12
Resolved
NO1H
6H
1D
1W
1M
ALL
e.g. Winograde >= 87.5%
This question is managed and resolved by Manifold.
Market context
Get 
1,000 to start trading!
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ118 | |
| 2 | Ṁ74 | |
| 3 | Ṁ45 | |
| 4 | Ṁ37 | |
| 5 | Ṁ31 |
People are also trading
Sort by:
predictedNO
the initial question was specifically about winogrande, but MMLU would also be acceptable. fits on a single 3090. quantization could be allowed if quantized performance is comparable, but that might be annoying to evaluate
Winograde? Do you mean the Winogrande benchmark? Like specifically this one? (Any requirement for the LLM in question to be open?)
https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard