| Model profile | ||
|---|---|---|
| Model | Size | Approximate size |
| Step 3.7 FlashModel source: provider262,144 tokens max native context | 198Bactive 11B | 118.8 GB4-bit quantization |
| Compatible hardware Estimated speeds, not benchmark results: calculated from memory bandwidth and model size. Real results can differ significantly because there is no precise formula for deriving LLM generation speed from hardware specifications alone. | ||||
|---|---|---|---|---|
| Hardware | Memory | Bandwidth | vLLM | Estimated generation |
| DGX Spark | 128 GBUnified RAM | 273 GB/s | yes | 16 tok/s4-bit quantization |
| M5 Max | 128 GBUnified RAM | 614 GB/s | no | 36 tok/s4-bit quantization |
| M5 Ultra | 256 GBUnified RAM | 1200 GB/s | no | 68 tok/s4-bit quantization |
| DGX H200 | 1128 GBGPU HBM3e | 4800 GB/s | yes | 209 tok/s4-bit quantization |
| DGX Station | 748 GBCoherent Memory | 7100 GB/s | yes | 269 tok/s4-bit quantization |
| ET900N G3 | 748 GBCoherent Memory | 7100 GB/s | yes | 269 tok/s4-bit quantization |
| DGX B200 | 1440 GBGPU HBM3e | 8000 GB/s | yes | 288 tok/s4-bit quantization |
| GB200 NVL72 | 13400 GBGPU HBM3e | 8000 GB/s | yes | 288 tok/s4-bit quantization |
| GB300 NVL72 | 20000 GBGPU HBM3e | 8000 GB/s | yes | 288 tok/s4-bit quantization |
Compares Qwen3.6 27B Q6 locally versus Step3.7 Flash via OpenRouter on speed, coding ability, creativity, interactivity, UI polish, and physics while building an iPhone-style UI, Sled Sketch, and Volt
Local testing of Step 3.7 Flash using Q4 and Q8 quantizations evaluates performance in browser OS, static scenes, games, coding, multimodal tasks, simulations, writing, and roleplay.