i still think llama4 maverick was a good model and y'all just used vllm and open router and didn't actually get to use the model that was trained. it was pretty close to gpt4o with better tool calling. not my favorite and was surpassed by qwen3 in tool calling but not terrible
10,88K