Hugging Face
Models
Datasets
Spaces
Buckets
new
Docs
Enterprise
Pricing
Website
Tasks
HuggingChat
Collections
Languages
Organizations
Community
Blog
Posts
Daily Papers
Hardware
Learn
Discord
Forum
GitHub
Solutions
Team & Enterprise
Hugging Face PRO
Enterprise Support
Inference Providers
Inference Endpoints
Storage Buckets
Log In
Sign Up
🤗
Scaling
214.2
TFLOPS
pavle
PRO
pavle-scalably
1
1
5
Follow
nadouch2002's profile picture
usamasharif143's profile picture
mayafree's profile picture
9 followers
·
24 following
https://scalably.io/author/pavle-lazic
pavlescalably
pavle-Scalably
pavle-lazic-scalably
AI & ML interests
Production AI agents, owned NVIDIA inference, frontier open-weight models, long-context serving, tool use, post-training and evaluation.
Recent Activity
replied
to
their
post
about 1 month ago
14 days serving https://huggingface.co/unsloth/Qwen3.8-27B-NVFP4 to production agents on 2x RTX 5090 (vLLM 0.27, TP=2, 262K context, FP8 KV): 28,097 requests, 860.6M prompt tokens, 82.6% prefix-cache hit rate, TTFT p50 0.61 s, 0 engine errors. The observation: prefix cache, not throughput, decides whether a 27B model keeps up with agents. Mean request is 30,100 tokens in, 983 out, because every turn resends the whole session. Two flags mattered most: --max-num-seqs 12 (queue p95 went 9.4 s to 233.6 s past that) and --watermark 0.08 (preemptions 29 to 2). And thinking off for tool loops: 917 tokens in 11.7 s vs 11,170 in 144 s, same answer. Full config and counters: scalably.io/blog/qwen3-8-27b-nvfp4-rtx-5090-production Next we are preparing an 8x B300 node in an EU data center for open-weight serving. Which models or workloads are underserved for you?
published
an
article
about 1 month ago
Fast memory, slow weights: how a production agent system improves around the model, and the model itself
replied
to
dipankarsarkar
's
post
about 1 month ago
Your issue tracker is in the wrong place. It lives on a server. Your code lives in git. So every time an agent picks up work it makes an API call, burns a token, fights a rate limit, and still cannot see what the other agent just did. Move the issues into the repo. Append-only event log in git refs. Branches when you branch, merges when you merge, CRDT so two agents never conflict. No server, no database. The coordination signal that PR-level telemetry misses lives before the pull request. The paper, and a live demo running the real tool: https://huggingface.co/papers/2606.19616 https://huggingface.co/spaces/neullabs/grite If your agents share a repo, where does their shared state actually live right now?
View all activity
Organizations
pavle-scalably
's activity
All
Models
Datasets
Spaces
Buckets
Papers
Collections
Community
Posts
Upvotes
Likes
Articles
liked
2 Spaces
about 1 month ago
Running
on
CPU Upgrade
Agents
567
LevelBot
🥇
567
Verify your Discord account with Hugging Face login
Disabled
Agents
1.74k
MiniMax H3 Turbo LoRA
🎬
1.74k
Video generation with a synchronized soundtrack
liked
3 models
about 1 month ago
Qwen/Qwen3.8-Flash-Next
Image-Text-to-Text
•
180B
•
Updated
Aug 27
•
2.53M
•
•
6.35k
unsloth/Qwen3.8-27B-GGUF
27B
•
Updated
Aug 20
•
6.41M
•
5.22k
Qwen/Qwen3.8-27B
Image-Text-to-Text
•
28B
•
Updated
Aug 14
•
7.59M
•
•
18k