LunarCrush LLM | creator/twitter::1145018230773362688/posts

[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

[@yangzhouy](/creator/twitter/yangzhouy)
"👀 Wanna run DeepSeek MoE models on AWS Cloud with DeepEP 1/ 🚀 Introducing UCCL-EP: A portable efficient Expert Parallelism framework that brings DeepEP-level GPU-driven communication with the same APIs to any cloud or hardware AWS EFA AMD GPUs Broadcom NICs and beyond"  
[X Link](https://x.com/yangzhouy/status/1982871914235474216) [@yangzhouy](/creator/x/yangzhouy) 2025-10-27T18:07Z XXX followers, XXX engagements


"@perplexity_ai wait our UCCL-ep already supports running MoE kernels on AWS with the same API as deepep. We released in last month: It already runs SGlang with deepseekV3 model. We also support amd GPUs + Broadcom NICs now. Our code is already open sourced"  
[X Link](https://x.com/yangzhouy/status/1986118875390632443) [@yangzhouy](/creator/x/yangzhouy) 2025-11-05T17:10Z XXX followers, XXX engagements


"Yes UCCL-ep has supported SGlang @sgl_project (without any line of code change) to run deepseekv3 on AWS H200 VMs. We are also integrating into vllm @vllm_project now. Will release soon UCCL-ep also supports amd + Broadcom low latency kernel now:"  
[X Link](https://x.com/yangzhouy/status/1986237537510449379) [@yangzhouy](/creator/x/yangzhouy) 2025-11-06T01:01Z XXX followers, 1141 engagements


"@AravSrinivas @AravSrinivas Our UCCL-ep supported MoE kernels on AWS EFA since last month: It also supports AMD GPUs and Broadcom NICs. We has better perf for prefill e.g. 2.1ms for dispatch 4.9ms for combine at EP32 while pplx has 4.7ms and 8.3ms respectively"  
[X Link](https://x.com/yangzhouy/status/1986288024947401131) [@yangzhouy](/creator/x/yangzhouy) 2025-11-06T04:22Z XXX followers, XX engagements

[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

@yangzhouy "👀 Wanna run DeepSeek MoE models on AWS Cloud with DeepEP 1/ 🚀 Introducing UCCL-EP: A portable efficient Expert Parallelism framework that brings DeepEP-level GPU-driven communication with the same APIs to any cloud or hardware AWS EFA AMD GPUs Broadcom NICs and beyond"
X Link @yangzhouy 2025-10-27T18:07Z XXX followers, XXX engagements

"@perplexity_ai wait our UCCL-ep already supports running MoE kernels on AWS with the same API as deepep. We released in last month: It already runs SGlang with deepseekV3 model. We also support amd GPUs + Broadcom NICs now. Our code is already open sourced"
X Link @yangzhouy 2025-11-05T17:10Z XXX followers, XXX engagements

"Yes UCCL-ep has supported SGlang @sgl_project (without any line of code change) to run deepseekv3 on AWS H200 VMs. We are also integrating into vllm @vllm_project now. Will release soon UCCL-ep also supports amd + Broadcom low latency kernel now:"
X Link @yangzhouy 2025-11-06T01:01Z XXX followers, 1141 engagements

"@AravSrinivas @AravSrinivas Our UCCL-ep supported MoE kernels on AWS EFA since last month: It also supports AMD GPUs and Broadcom NICs. We has better perf for prefill e.g. 2.1ms for dispatch 4.9ms for combine at EP32 while pplx has 4.7ms and 8.3ms respectively"
X Link @yangzhouy 2025-11-06T04:22Z XXX followers, XX engagements