1T/32B-active MoE with 256K context and MoonViT vision, Modified MIT license; native multimodal with agent-swarm parallel sub-task orchestration.
@hn_c64414
about 1 month ago
How do people evaluate creative writing and emotional intelligence in LLMs? Most benchmarks seem to focus on reasoning or correctness, which feels orthogonal. I’ve been playing with Kimmy K 2.5 and it feels much stronger on voice and emotional grounding, but I don’t know how to measure that beyond human judgment.
@rd_3994b9
2 months ago
Sometimes default to kimi 2.5 and that range of models (I know composer is a better kimi 2.5)
@hn_af0219
2 months ago
If it helps, I mean it in a really literal sense. qwen3.6 27b is currently $3.20 per million tokens on openrouter right now which is way overpriced. As good as the 27b is, kimi k2.5 $3.00 and it's just in another league in terms of capability. There's no reason to spend money on it. And even alibaba's own qwen3.6-plus is $1.95, so it's kinda easy to come to a conclusion that alibaba (nor anyone else) is really interested in hosting that model. And don't get me wrong, I fully agree with you, qwen3.6 27b is an amazing model. I run it on my own hardware and every day I'm constantly surprised with what it can zero shot.
@hn_e3e563
3 months ago
70B dense models are way behind SOTA. Even the aforementioned Kimi 2.5 has fewer active parameters than that, and then quantized at int4. We're at a point where some near-frontier models may run out of the box on Mac Mini-grade hardware, with perhaps no real need to even upgrade to the Mac Studio.