192 NVIDIA B300 GPUs running DeepSeek V3 gained 1.13% throughput with a single software tweak.
One sentence → playable game.
Open-source plugin for DeepSeek Harness. dsh-game-studio by FiniYang: prompt → 2D or 3D voxel game (Three.js) → live preview in the chat sidebar → agent playtests it headless: fps, errors, screenshots → swap code-drawn characters for AI art →
Choose your AI assistant, Little Guga or DeepSeek.
🤖
'Small' here means compute, not memory: 17B of 594B is under 3% active per token
But all 594B still have to sit in memory, about 1.2 TB in BF16 or roughly 300 GB at 4 bits. The Apache 2.0 licence is the other big deal, given today's ERNIE release excludes whole regions. How MoE
You have reached the end of the archive
All of deepseek