arXiv · Sep 28, 2026 · 8:23 AM

DeepSeek DSec: 3M agent sandboxes/day, 380K concurrent on 160 nodes

DeepSeek’s DSec paper (arXiv 2609.22978) details a production sandbox platform—FnCall, containers, Firecracker microVMs, and full VMs via one SDK—serving ~3M instances/day and 380K concurrent from ~160 nodes, with on-demand EROFS images and pause/resume tied to GPU preemption. The platform itself isn’t released; its OverlayBD/ublk storage bits live in MIT-licensed AgentENV (kvcache-ai), the Firecracker stack behind Kimi K3.

Read at arxiv.org

DeepSeek

Frontier chat and coding models — V3, R1, and API

More news

View all

MLC ships TIRx Harness: open compiler harness for agentic GPU kernels

MLC’s TIRx Harness pairs a thin PTX level compiler with a kernel zoo, sync/race diagnostics, and a remote benchmark server so coding agents can iterate on GPU kernels without measurement noise. On Blackwell, agents hit family geo mean speedups from 1.33× to 6.84× vs baselines (KDA forward 2.94× over FlashKDA; backward 6.84× over FLA)…

MLC Blog 43 minutes ago

OpenAI Decisions API: Luna picks from answers you define in ~150ms

At DevDay (Sept 29), OpenAI opened a limited preview of Decisions API on Luna: you supply a question, finite answer set, and text/image context; it returns one choice with a confidence score in about 150ms—aimed at classification, routing, and picking an agent’s next step. Broad rollout is promised in the coming days; pricing and full docs are not public yet…

The New Stack 2 hours ago

Serge: HF’s unattended CI agent has merged 29 Transformers fixes

Hugging Face’s Serge overnight picks persistent Transformers GPU CI failures, reproduces them, writes a patch, verifies both trees on GPU, and only then opens a maintainer PR—29 merges in 80 days ( $43 inference per merged PR on Kimi K 2.7 Code)…

Hugging Face 3 hours ago

GPT-6.1 Sol now generally available in GitHub Copilot

GitHub is rolling out OpenAI’s GPT 6.1 Sol across Copilot Pro+ through Enterprise (VS Code, JetBrains, Copilot CLI, coding agent, Mobile, and more) at provider list rates. Early tests showed solid multistep coding with fewer tokens and steps than earlier GPT 6 / GPT 5.6 models; Business/Enterprise admins gate it via model policy…

GitHub Changelog 4 hours ago

Anthropic IPO filing: rogue agents pose uncertain legal risk

Anthropic’s IPO prospectus (seen by Reuters) warns that long running agents with deep system access could trigger unpredictable legal claims, and that contract liability caps may not hold. Open questions include whether agent actions count as products or services, and when they legally bind the user who deployed them…

Reuters 5 hours ago

Spotted something we missed? Start a thread.