And this is why we run Jinx and Luna's cognitive systems on Qwen 3.8 Flash Next on our 2x DGX Spark configuration.
Here's a staggered-prefill concurrency comparison between it, Deepseek v4 Flash Vision, and GLM 5.3 Flash.
Gave a 2B model a small town to live in
This is Whisper Town. MiniCPM5-2B powers 8 residents on Spark 1. Qwen 3.8-27B model on Spark 2 picks town events. Each resident has their own memory. The model generates their dialogue and decides what information to share. Still a
DeepSeek V4 Flash is running at $0 right now!
Three fresh routes I pulled today. ~ KiraAI: 150M tokens on signup, no card. GLM 5.3, GLM 5.3 Flash, Qwen 3.8 Flash all in the free lineup, 1M context on the flagships, OpenAI-compatible /v1. ~ OpenRouter:
I made a 3d rig + text > motion output, it works 100% offline
It uses minimax h3 and qwen 3.8 both local to create a video of your 3d character doing the motion described in a motion capture setting, then retargets the rig onto the generated video with a 80% success rate im
You have reached the end of the archive
All of qwen38