mini-infer - PagedAttention KV-Cache & Prefix Studio
Interactive GPU PagedAttention block pool & prefix cache.
Interactive GPU PagedAttention block pool & prefix cache.
Discrete-event LLM scheduler simulator & Gantt player.
Sub-byte quantization Pareto frontier & BitNet packer.
Pre-flight LLM VRAM sizing CAD & cloud TCO planner.
Empirical consumer GPU serving benchmarks on Dual T4 & 3050.
Compare how different LLM tokenizers split your text
Visualize speculative LLM streaming performance