Running Qwen3.8 at 255 tok/s on a single 5090

Automated excerpt

Task-specialized open models for your inference needs. We take the latest open source models, and through a series of post-training, quantization and kernel specialization techniques, we make them run fast on local hardware. DeepSeek-V4. 1-Flash-Ternary- Recommended hardware: 4x RTX6000- 4 users, 276 tok/sec aggregate- for complex coding and researchQwen3. 8-Flash-Next-NVFP4- Recommended hardware: 1x RTX6000- 4 users, 334 tok/sec aggregate- for coding and everyday taskswith the Goethe framework:~ git clone https://github. com/meshapplied/goethe.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.