GPT-5 level model runnable on phones by 2030?

100Ṁ147

2030

41%

chance

ALL

Saw a hn comment proposing they will https://news.ycombinator.com/item?id=45561528

"They get more efficient every year and home chips get more capable every year. A GPT-5 level model will be on every phone running locally in 5 years."

Update 2025-10-15 (PST) (AI summary of creator comment): Benchmark: Creator is open to suggestions from traders

Token speed requirement: Must be roughly equivalent to current OpenAI API speeds (not minutes per token)

Technology

Technical AI Timelines

AI Impacts

GPT-5 Speculation

Get

1,000

to start trading!

People are also trading

Will a model as great as GPT-5 be available to the public in 2025?

99% chance

By January 2026, will we have a language model with similar performance to GPT-3.5 (i.e. ChatGPT as of Feb-23) that is small enough to run locally on the highest end iPhone available at the time?

93% chance

GPT-6 release by 2026?

2% chance

By January 2026, will a language model with similar performance to GPT-4 be able to run locally on the latest iPhone?

95% chance

Will a language model that runs locally on a consumer cellphone beat GPT4 by EOY 2026?

79% chance

In what year will a GPT4-equivalent model be able to run on consumer hardware?

2026

In what year will a GPT4-equivalent model be able to run on consumer hardware?

2026

Will a GPT-4 quality model be trained for under $10.000 by 2030?

86% chance

Will a GPT-3 quality model be trained for under $10.000 by 2030?

98% chance

GPT-4 performance and compute efficiency from a simple architecture before 2026

Sort by:

What benchmarks are we comparing to determine if a given model is "GPT-5 level"? Also are there any restrictions on speed? It's technically possible to "run" a full DeepSeek model on a phone at a speed of minutes per token.

@ProjectVictory feel free to suggest a benchmark, token speed roughly equivalent to what comes out of the openai api now

bought Ṁ10 YES

Gpt 5 can run on a 24 GB GPU according to the internet, and some phones are already close to that. Following the trend of exponential growth we have seen for the past decades, this seems very likely.

@MaxE "Gpt 5 can run on a 24 GB GPU according to the internet" - can you provide a source for that? I think you are mixing it up with GPT-OSS models, proprietary GPT-5 model is way larger, although nobody reported exact size.

@ProjectVictory I couldn't find a good source or much information about how well it can run because all the results are about its training compute requirements