|
| ▲ | kakugawa 30 minutes ago | parent | next [-] |
| Yeah, a good balance would be a local reasoning model paired w/ a knowledge pack + web-based searches. You want your obscure coding language? Provide a pre-vetted knowledge pack locally that it knows how to use. Same with any other domain-specific knowledge. |
| |
| ▲ | wolvoleo 27 minutes ago | parent [-] | | Yes especially now that local context isn't so costly anymore. With llama3 it would kill my vram just to have a few thousand tokens. With recent models it's hardly an impact anymore. Not sure how they manage that but it's great. It still takes more processing but I also run an older card that is not very good at prefill. |
|
|
| ▲ | InsideOutSanta 23 minutes ago | parent | prev | next [-] |
| > Locally you're about a year or so behind SOTA If that holds, I'll be able to do 99% of my work with local models around March 2027. |
|
| ▲ | furyofantares 29 minutes ago | parent | prev [-] |
| Would local models catch up if frontier models paused new training for a year? I'm not sure for. It might just be you can get these nice local models because of distillation from a much more powerful model. |
| |
| ▲ | swiftcoder 21 minutes ago | parent [-] | | Adversarial distillation hasn’t been a major factor in the past year. If the SOTA stops moving, all the other labs catch up to them in 6 months to a year, and distillation into smaller models continues to close the gap between small/large models during that time |
|