ResearchLMSYS·5d ago·3 min readSummaryListen 0:48
XM-Sparse: Pushing the Throughput Frontier of Sparse Attention with Netpreme X-Mem™XM-Sparse improves sparse attention throughput using Netpreme X-Mem technology.
News Hub
What actually shipped in agent engineering, pulled from the labs, arXiv and Hacker News.
See who we follow59 articles
ResearchLMSYS·8d ago·3 min readSummaryListen 0:43
HiCache Across Multi-Parallelism and Distributed ServingHiCache optimizes multi-parallelism and distributed serving for language models.
ResearchLMSYS·16d ago·3 min readSummaryListen 1:02
miniTE Technical Report: Stable and Efficient Reinforcement Learning with Native NVFP4LMSYS published miniTE technical report on reinforcement learning with NVFP4.
ResearchLMSYS·19d ago·3 min readSummaryListen 0:45
Training SWE agents with Miles on MI355X: taming long-tail trajectories and complex sandbox coordinationLMSYS trained software engineering agents using Miles on MI355X GPUs for complex sandbox coordination.
IndustryLMSYS·19d ago·3 min readSummaryListen 0:47
Announcing the SGLang AWS Trainium Plugin BetaSGLang AWS Trainium Plugin Beta announced.
ResearchLMSYS·21d ago·3 min readSummaryListen 0:42
SGLang-TorchTPU: PyTorch-Native Production Serving on TPU, Built with Google and MetaSGLang-TorchTPU enables PyTorch-native production serving on TPU, built collaboratively by Google and Meta.
ResearchLMSYS·23d ago·3 min readSummaryListen 0:36
SGLang Low-Latency Mode: Kernels and RuntimeSGLang Low-Latency Mode optimizes kernels and runtime for faster inference.
ResearchLMSYS·55d ago·3 min readSummaryListen 0:42
Accelerating Long-Context and Agentic Inference with NVFP4 KV CacheNVFP4 KV cache accelerates long-context and agentic inference.
ResearchLMSYS·61d ago·3 min readSummaryListen 0:53
SGLang and Miles Add Day-0 Support for DeepSeek-V4.1SGLang and Miles add day-zero support for DeepSeek-V4.1.
ResearchLMSYS·73d ago·3 min read
Running DeepSeek-V4-Flash and Kimi-K3 on Consumer Hardware with SSD Expert PackDeepSeek-V4-Flash and Kimi-K3 models can run on consumer hardware using SSD Expert Pack.
ResearchLMSYS·75d ago·3 min readSummary
MiniMax-H3 on 8×H200: 1.95× Lossless, Up to 6.24× at 0.76MiniMax-H3 achieves 1.95x lossless speedup on 8 H200 GPUs, scaling to 6.24x at reduced precision.
ResearchLMSYS·76d ago·3 min readSummary
Qwen3.8-Flash-Next: Day-0 Support in SGLangSGLang provides day-zero support for Qwen3.8-Flash-Next model.
ResearchLMSYS·81d ago·3 min readSummaryListen 0:51
Chasing the Batch-1 Floor: Ling-3.0-flash Speculative Decode on BlackwellLing-3.0-flash uses speculative decoding on Blackwell to optimize batch-1 inference performance.
ResearchLMSYS·82d ago·3 min readSummaryListen 0:45
Mooncake for Miles: From Fragmented Rollout Data to Efficient Bulk I/OMooncake converts fragmented rollout data into efficient bulk I/O operations.
ResearchLMSYS·83d ago·3 min readSummaryListen 0:47
Pushing the Limits of Serving DeepSeek-V4-ProLMSYS examined serving performance limits for DeepSeek-V4-Pro models.
ResearchLMSYS·84d ago·3 min readSummaryListen 0:47
Miles v0.1: Production-level Post-trainingLMSYS released Miles v0.1 for production-level post-training of language models.
ResearchLMSYS·85d ago·3 min readSummaryListen 0:48
Advanced CUDA Graph Techniques in InferenceAdvanced CUDA Graph Techniques in Inference discusses optimization methods for model inference.
ResearchLMSYS·90d ago·3 min readSummaryListen 0:52
SGLang and Miles Add Day-0 Support for Qwen3.8SGLang and Miles add day-0 support for Qwen3.8 model.
ResearchLMSYS·91d ago·3 min readSummaryListen 0:45
SGLang Adds Day-0 Support for NVIDIA Nemotron 3.5 LightningSGLang added immediate support for NVIDIA Nemotron 3.5 Lightning.
ResearchLMSYS·92d ago·3 min readSummaryListen 0:51
SGLang Adds Day-0 Support for Muse Glimmer, a Multimodal Model Built for Local Agentic WorkflowsSGLang added immediate support for Muse Glimmer, a multimodal model designed for local agentic workflows.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.