Qwen3.8-27B on RTX 3090
Inference stack for Qwen3.8-27B on a single RTX 3090: 195k context, 82 tps single request, 672 tps peak.
Best Qwen Builds on X
Qwen 3.8 and earlier. Local models that actually wrote a game, a CLI or a bench you can clone. 8 builds on this page, each with the original post.
Qwen3.8-27B NVFP4 on DGX Spark
Day-one Qwen3.8-27B deploy: NVFP4, 262K context, one DGX Spark, one command. Full reproducible stack.
Qwen3.8-27B FP8 on 2x 3090
27B on two consumer 3090s at ~75-85 tok/s with NVLink. Tool calling, reasoning, vision, MTP. Full deploy open-sourced.
dupe
Duplicate-file finder one-shotted by local Qwen 3.8 27B on a 128GB Mac via OpenCode.
Hermes Builder Doctor
Hermes builder tools updated after running Qwen 3.8 27B so smaller local models can break work into parts.
Starskip
Game built and debugged with a DeepSeek V4 Pro / Qwen distill in Claude Code. Live on Hugging Face Spaces.
Tetris Battle
Same prompt, six models: DeepSeek V4 Flash, Kimi K3, Qwen 3.8, GLM and Grok. Playable results plus repo.
Local Qwen Breakout
A single-file HTML canvas Breakout game written end to end by a local Qwen 3.8 27B through Hermes.
