#扩散模型
3 related articles
Tech Frontiers·2 min
Mercury 2: How Diffusion Architecture Achieves 1000 Tokens Per Second Inference Speed
Inception Labs releases Mercury 2, a diffusion-based language model achieving 1000+ tokens/sec with strong reasoning, challenging autoregressive architectures.
Read more →
Tech Frontiers·2 min
Mercury 2 Hands-On Review: The Fastest Inference LLM Powered by Diffusion Models — A Complete Game Generated in 18 Seconds
Hands-on review of Inception Labs' Mercury 2 diffusion model, benchmarked against Claude Haiku, Gemini Flash and more across code generation, structured reasoning, and long-range planning at 1000+ tokens/sec.
Read more →
Research·3 min
SVDQuant: 4-bit Quantization Enables Efficient Diffusion Model Inference on Consumer GPUs
SVDQuant, an ICLR 2025 Spotlight paper, achieves 4-bit diffusion model quantization via low-rank decomposition that absorbs outliers, reducing memory by 75%. Open-source engine Nunchaku (3800+ stars) enables FLUX inference on consumer GPUs like RTX 4060.
Read more →