Prefill Decoding

Feeds to Scour
SubscribedAll
Scoured 179 posts in 9.6 ms

RKSC: Reasoning-Aware KV Cache Sharing and Confident Early Exit for Multi-Step LLM Inference

 💾KV Cache  Content type: Academic
arxiv.org·
Less-relevant results

A system programmer’s guide to LLM inference

 💰Inference Cost  Content type: Blog

2x GH200 for LLM inference, Part 2: vLLM, DeepSeek V4 Flash, and MTP

 🧠Inference Engineering  Content type: Blog
dnhkng.github.io·

LangChain vs LlamaIndex 2026: Response Time on 10 RAG Tasks

 ⚙️ML Compilers  Content type: Blog  Content type: Discussion
tildalice.io·

Two Leaps to 1000 Tokens/s on a 1T-Parameter Model: On Inference Systems, Execution Boundaries, and Co-Design

 🧠Inference Engineering  Content type: Blog
tilert.ai··Hacker News

MLPerf and the rise of latency-aware LLM benchmarking

 🧠Inference Engineering
edn.com·

KaiFelixBennett/gemma4-turboquant-rdna4: Run Gemma-4-31B at full 256K context on a $1,400 AMD RDNA4 GPU (gfx1201): TurboQuant KV cache + HIP-graph-safe Flash-Attention for llama.cpp, fully measured on real hardware.

 FlashAttention  Content type: Code
github.com··Hacker News

The economics of speculative decoding

 🚀Speculative Decoding  Content type: Blog

Running LLM Inference on Kubernetes: What It Actually Takes

 🧠Inference Engineering  Content type: Blog
fairwinds.com·

From GPU to Token: The 8-Layer Observability Stack for AI Infrastructure

 💰Inference Cost  Content type: Blog
jimmysong.io·

ReasonAlloc: Hierarchical Decoding-Time KV Cache Budget Allocation for Reasoning Models

 🧠Inference Engineering  Content type: Academic
arxiv.org·

Massive AI Storage Demand Creates a New Memory Wall

 🧠Inference Engineering  Content type: News
eetimes.com·

The Death of the Four Golden Signals: Designing Telemetry for Non-Deterministic Infrastructure

 🛡️SRE
devops.com·

AMD's Lemonade SDK For Local AI Adds NVIDIA CUDA Support

 🎮GPU Computing
phoronix.com·

Which is faster: Gemini 3.5 Flash or Kimi K2.6 on Cerebras

 🔢FP8 Training  Content type: Blog
cerebras.ai·

Benchmarking dots.tts on Strix Halo

 🎮GPU Computing
sleepingrobots.com·

What Breaks When Multi-Agent Systems Scale

 🛡️SRE
digitalocean.com·

Breaking the Ice: Analyzing Cold Start Latency in vLLM

 💾KV Cache  Content type: Academic
arxiv.org··Hacker News

Nvidia DGX Spark GB10 – AI Models and Guide with vLLM and Autonomous Script

 💾KV Cache  Content type: Code
github.com··Hacker News

Cohere open-sources a coding agent that runs on a single H100

 🎮GPU Computing
venturebeat.com·

Keyboard Shortcuts

Navigation

Next / previous item
j/k
Open post
oorEnter
Preview post
v

Post Actions

Love post
a
Like post
l
Dislike post
d
Undo reaction
u
Save / unsave
s

Recommendations

Add interest / feed
Enter
Not interested
x

Go to

Home
gh
Interests
gi
Feeds
gf
Likes
gl
History
gy
Changelog
gc
Settings
gs
Browse
gb
Search
/

General

Show this help
?
Submit feedback
!
Close modal / unfocus
Esc

Press ? anytime to show this help