3.0
@hn_720bcc
about 1 month ago
I think this is old news, but this model does better than llama 4 maverick on coding.
Meta's Llama 4 Maverick MoE model with 17B active / 128 experts (400B total), 1M token context, natively multimodal, under the Llama 4 Community License.
@hn_720bcc
about 1 month ago
I think this is old news, but this model does better than llama 4 maverick on coding.
@hn_79b55c
about 2 months ago
Exactly, it's all server side. There are no plans for this. The main issue I see with doing it in device is the LLM piece. Even with some large models like llama 4 maverick, the tutor just struggles to properly teach and understand the student, it's not viable IMO. Intelligence is super key here, especially as the context size gets larger (due to memory) and intelligence degrades. Another major issue is TTS voice quality, but this seems to be improving a lot for small local models. EDIT: You're right, latency is also a big deal. You need to get each piece under a second, and the LLM part would be especially slow on mobile devices.
@hn_a65d26
3 months ago
I’m a newish Kagi user and I find myself using the LLM about as frequently as search itself. Sometimes I search for things I know I am looking for. Other times I don’t know quite what I’m looking for or I know in advance that I’m not likely to find it—so I chuck it at Llama 4 Maverick and it usually gives me something useful. I had no plans to use the LLMs until they opened it up on my tier. At this point however, it’s half the value I get out of Kagi.