Will an LLM better than gpt3.5 run on my rtx 3090 before 2025?

1kṀ2002

resolved Jan 2

Resolved

YES

ALL

Inspired by these questions:

/sylv/an-llm-as-capable-as-gpt4-will-run-f290970e1a03 ‌

/sylv/an-llm-as-capable-as-gpt4-runs-on-o

Resolution criteria (provisional):

Same as /singer/will-an-llm-better-than-gpt4-run-on, but replace "gpt4" with "gpt3.5".

Update 2025-01-01 (PST) (AI summary of creator comment): - Resolution will be based on the creator's personal judgment to save time.
- If you disagree with the resolution, please DM the creator for a refund.

Technology

Get

1,000

to start trading!

🏅 Top traders

#	Name	Total profit
1		Ṁ60
2		Ṁ52
3		Ṁ25
4		Ṁ17
5		Ṁ4

People are also trading

China will make a LLM approximately as good or better than GPT4 before 2025

89% chance

Will a LLM trained with FP4 have frontier-level performance before 2028?

31% chance

What LLM API will work unmodified until the end of 2025?

Size of smallest open-source LLM marching GPT 3.5's performance in 2025? (GB)

1.83

When will an open-source LLM be released with a better performance than GPT-4?

How much time will pass between an LLM being released that beats GPT4 and the next OpenAI LLM being released? (+ANSWERS)

Will there be a state-of-the-art LLM that is NOT based on next raw token prediction before 2029?

Sort by:

@traders I'm resolving this based on personal judgment to save time. If you disagree, please DM me and I'll refund you.

@singer presumably this is deepseek? I think you shouldn't refund people who disagree with you but that's just me

@Bayesian it's not any model in particular. it's been so long since I used gpt3.5 that I really can't remember how good it was compared to the local models I've been using. I offered to refund because I resolved based on vibes

Note (I'm adding this comment to a few of my markets): I was hoping to do regular early tests of this but it's too far back on my backlog right now. I'm still committing to resolving this properly at the end of the year, however.

Mixtral has just pulled ahead of gpt3.5. When I have time I'll see if I can find a quantized model I can run that doesn't do more than 2% worse on the Winograd Schema Challenge.