@droidbuilds Mythos will use all of your tokens just to determine you shouldn’t use ai and should go outside and touch grass without ever answering the prompt. GPT is just going to do the work.
@itsPaulAi@editxshub But connecting it to the web is a tool call and both gemini and gemma struggle with tool calls. They are great chat models, but this shortcoming is why people fall back to qwen.
@Pallavi_345 It’s not a good model, doesn’t even have access to search google’s other services like youtube. It also hallucinates more than any other frontier model.
@brettrigby7@davideciffa System memory is the only way you’re going to fit useful models… This is slower than last year’s Macbook pro and it’s not even out yet.
@shiri_shh This is a slower version of the same chip in the DGX spark which was already slow at release last year. This chip has been delayed a year with no performance improvement.
@toddsaunders Opus 4.7 was all sorts of broken. It couldn’t do simple things like place real values in a four line template it created without hallucinating similar values. Opus 4.8 when pointed to the same conversation called out Opus 4.7 on its shit responses.
@TheGameVerse This is horrible, flat textures, focus on reflectivity when it’s not realistic for the vehicle, bad physics, water reflectivity fake, twenty year old tech here.