---
type: "post"
id: "vercel-sandboxes-with-drives-start-2-4x-faster-4873e5"
slug: "vercel-sandboxes-with-drives-start-2-4x-faster-4873e5"
title: "Vercel Sandboxes with Drives start 2.4x faster."
description: "Vercel says Sandboxes that mount Drives now start faster: median creation time fell from 1.7 seconds to 0.7 seconds, while p95 fell from 2.2 seconds to 1.6 seconds."
retrieval_nugget: "Vercel says Sandboxes that mount Drives now start faster: median creation time fell from 1.7 seconds to 0.7 seconds, while p95 fell from 2.2 seconds to 1.6 seconds. The change is automatic, needs no SDK or configuration update, and leaves pricing unchanged. It matters where teams repeatedly create these environments."
published_at: "2026-11-02"
updated_at: "2026-11-02"
record_date: "2026-11-02"
date_kind: "published_at"
topics: ["sandbox"]
entities: ["Vercel","Vercel Sandbox","Vercel Drives"]
related_patterns: ["agent-ready-software-exposes-capabilities"]
novelty_label: "new_fact"
what_changed: "Vercel reports Drive-enabled Sandbox startup is up to 2.4× faster: median creation time fell from 1.7s to 0.7s and p95 from 2.2s to 1.6s, with no configuration changes or pricing impact."
prior_refs: []
source_url: "https://vercel.com/changelog/vercel-sandbox-with-drives-now-starts-up-to-2-4x-faster"
source_urls: ["https://vercel.com/changelog/vercel-sandbox-with-drives-now-starts-up-to-2-4x-faster"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:vercel-sandboxes-with-drives-start-2-4x-faster-4873e5"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/vercel-sandboxes-with-drives-start-2-4x-faster-4873e5/","markdown":"https://newruntime.com/posts/vercel-sandboxes-with-drives-start-2-4x-faster-4873e5.md","json":"https://newruntime.com/posts/vercel-sandboxes-with-drives-start-2-4x-faster-4873e5.json"}
markdown_url: "https://newruntime.com/posts/vercel-sandboxes-with-drives-start-2-4x-faster-4873e5.md"
---

# Vercel Sandboxes with Drives start 2.4x faster.

## Retrieval answer

Vercel says Sandboxes that mount Drives now start faster: median creation time fell from 1.7 seconds to 0.7 seconds, while p95 fell from 2.2 seconds to 1.6 seconds. The change is automatic, needs no SDK or configuration update, and leaves pricing unchanged. It matters where teams repeatedly create these environments.

## Summary
Vercel says Sandboxes that mount Drives now start faster: median creation time fell from 1.7 seconds to 0.7 seconds, while p95 fell from 2.2 seconds to 1.6 seconds. The change is automatic, needs no SDK or configuration update, and leaves pricing unchanged. It matters where teams repeatedly create these environments.

## What happened
Vercel has published a performance update for Vercel Sandboxes that mount Drives. The company says the startup improvement applies to the creation of a sandbox with Drives from a Vercel image. Its changelog characterizes the improvement as startup “up to 2.4x faster.” The supporting before-and-after comparison gives two creation-time measurements rather than a single aggregate figure.

The median time to create one of these sandboxes fell from 1.7 seconds to 0.7 seconds. That is a reduction of one second at the median, and is the basis for the stated up-to-2.4x improvement. At the 95th percentile, Vercel reports that creation time fell from 2.2 seconds to 1.6 seconds. The p95 result is therefore a smaller absolute reduction of 0.6 seconds, but still indicates an improvement in the slower end of the reported creation-time distribution.

The update is not presented as a new product tier, an API addition, or a change to how a developer configures a sandbox. Vercel says the speedup applies automatically to every sandbox that uses Drives. Existing users do not need an SDK upgrade, configuration change, or other migration step to receive it. The company also says that pricing stays the same, so the announcement does not introduce a stated price or access trade-off for the faster startup.

The scope matters. The source refers to creating a Vercel Sandbox with Drives, specifically from a Vercel image. It does not claim that every Vercel Sandbox starts faster, that sandbox commands execute faster after startup, or that a complete agent task will finish 2.4 times sooner. Nor does it give a separate result for sandboxes without Drives. The announcement links to Vercel’s Drives documentation, but the supplied material does not describe the implementation behind the improvement.

For a workflow that creates environments frequently, creation latency is a distinct part of the path before any application command, agent loop, or tool call can begin. Vercel’s announcement supplies a narrower claim: the time to create the relevant sandbox has improved, with median and p95 figures supplied for that operation. It does not offer a broader end-to-end benchmark.

## What the archive adds
This is a new performance fact in a continuing Vercel Sandbox thread. The earlier Field Note, [Vercel Sandbox commands now accept stdin, enabling stdio MCP servers inside sandboxes](https://newruntime.com/posts/vercel-sandbox-commands-now-accept-stdin-enabling-stdio-d16cc1/), covered interactive command input and stdio MCP servers. That was a capability change; this announcement adds a startup metric rather than changing command behavior.

## Trend signal
This weakly supports [agent-ready software exposes capabilities](https://newruntime.com/patterns/agent-ready-software-exposes-capabilities). A managed sandbox is a runtime surface through which teams execute bounded work, but this release adds no observable or permissioned capability. It is evidence about the performance of an existing runtime surface, not strong evidence for the pattern’s central claim.

## Why now
Vercel provides before-and-after creation-time figures and says the change is already automatic for Drive-using sandboxes. It does not identify a customer trigger, an underlying technical change, or a separate launch date that explains the timing of the announcement.

## Why it matters
For teams that create many Vercel Sandboxes with Drives, a one-second reduction at the median can reduce accumulated waiting around short, repeated, or parallel jobs. The p95 change also reduces the reported delay for slower creations. That can improve the responsiveness of a workflow whose next step cannot begin until an isolated environment exists.

The operational benefit is unusually low-friction: teams already using Drives do not need to change SDK versions, modify configuration, or accept a stated pricing increase. Still, creation time is only one component of an agent workflow. Repository setup, dependency installation, tool calls, model latency, command execution, review, and retries may dominate overall completion time. Teams should measure their own startup and completion path before treating Vercel’s creation-time result as an end-to-end productivity gain.

## What happens next
- **Teams with frequent Drive-backed sandbox creation will compare their own startup latency against the published median and p95 figures.** (3 months, 65%). Watch for: workload-level benchmarks or customer reports.
- **Vercel may publish further sandbox performance measurements or implementation details.** (6 months, 35%). Watch for: changelog updates that expand the scope beyond Drive-backed creation.

## Limits and uncertainty
The announcement does not provide sample size, test period, region, hardware details, workload variance, or a detailed definition of the population used for its p95 measure. It does not explain what changed internally, and it does not establish the same gains for sandboxes without Drives. The numbers concern creation from a Vercel image, not command execution, compute throughput, or total task completion time. Vercel says pricing is unchanged, but supplies no pricing schedule in this update. The source also does not state when the performance change took effect.

## Skill to build now
Instrument the interval from a request to create a sandbox until the first useful command can run, then record it separately from command and model time. For a Drive-backed Vercel Sandbox workflow, compare median and tail startup latency before and after normal deployments. This gives a team a way to determine whether environment creation is a material bottleneck rather than assuming a platform-level creation-time improvement changes total job duration.

## Reader takeaway
If your team already uses Vercel Sandboxes with Drives, the reported startup improvement is automatic and requires no SDK or configuration work. Treat it as a targeted creation-time gain: it may reduce waiting before a job begins, but it is not evidence that every workload or agent task will complete faster.

## Open question
What change produced the reported startup improvement, and do the same gains hold across different regions, image types, and workload sizes?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Sandbox](https://newruntime.com/topics/sandbox.md): Explore the sandbox topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)