Benchmarks
Measured 27 Oct 2026 from a client in us-east-1, on the same machines that serve this site. Every figure here is read from the raw samples, which are public.
475 ms
a whole session on Browser Arena
6th of 13 by the median, 4th by p95 (564 ms)
222 ms
new browser, request → answering CDP
p95 273 ms
223 ms
paused while it moves machines
a live client’s longest wait: 595 ms
201 ms
paused to fork a running browser
3 copies ready in 2.6 s
Browser Arena
The harness behind browserarena.ai with Baton added as a provider and nothing else changed: create, connect over CDP, load example.com, release. 10 warm-ups, then 100 sessions one at a time, from a t3.small in us-east-1, on a free account.
Browser Arena
Milliseconds per session, lower is better
- N
Notte
340p50291p99430 
Browser Use
490p50424p99639- L
Lightpanda
559p50485p99674 Baton
564p50475p99630Browserbase
663p50590p99730Kernel (headful)
676p50344p99767Kernel
712p50378p99741
Tilion
922p50156p9913,292
Hyperbrowser
975p50890p991,280
Cloudflare
1,546p50962p991,919Steel
1,691p501,293p992,577
Anchor
3,164p502,552p993,937
AWS AgentCore
3,627p502,613p994,411
Every other row is Browser Arena’s own latest run, from the raw results it publishes; we did not re-run them. Baton’s row is our run of the same harness from t3.small in us-east-1, on a free account, 100 of 100 sessions completed.
One day of work on it, 2026-10-24
the sum of the stage medians, as Browser Arena reports it
| create | connect | load page | release | total | p95 | |
|---|---|---|---|---|---|---|
| Start of the day | 244 ms | 259 ms | 489 ms | 206 ms | 1,198 ms | 1,331 ms |
| Release in the background | 232 ms | 262 ms | 493 ms | 51 ms | 1,038 ms | 1,148 ms |
| Warm pool | 70 ms | 122 ms | 323 ms | 53 ms | 568 ms | 688 ms |
| One-hop API | 58 ms | 99 ms | 270 ms | 43 ms | 470 ms | 564 ms |
Where the rest goes, and why we stopped here
- Sessions start from a browser that is already running and warmed up, a pool of one per machine; claiming one is a database update. That took create from 244 ms to 58 ms and connect from 259 ms to 99 ms.
- Loading the page is most of what is left, and it is Chrome, not the network: our gateway fetches example.com in about 25 ms. A browser in a VM on nested virtualization pays for every first touch of a memory page (12–17 µs measured) and every wake-up across CPUs. The fastest provider here loads the page in about 46 ms.
- About 40 ms of connect is CloudFront, which gives us TLS without a domain of our own.
- Getting near 150 ms would take bare-metal hosts (EC2 bare metal starts around $4 an hour per machine, always on) and a TLS endpoint of our own. That is spending, not engineering, so we stopped.
- The pool’s costs: one idle browser per machine, and a warmed browser has touched more memory, so pausing one stores about 740 MiB instead of 610 (pause 2.2 s instead of 1.9).
Steel’s browserbench
Steel’s open-source benchmark with Baton added: create, connect, load google.com, release; 300 runs (300 succeeded) from a t3.small in us-east-1 on October 24, before the warm pool and the one-hop API. The other rows are the numbers Steel published; we did not re-run them.
Milliseconds per session
total is the median; stages are averages, as in Steel’s table
| total p50 | p95 | create | connect | load page | release | all but page | |
|---|---|---|---|---|---|---|---|
| Steel | 867 ms | 1,090 ms | 182 ms | 175 ms | 490 ms | 48 ms | 404 ms |
| Kernel | 1,485 ms | 1,752 ms | 261 ms | 429 ms | 534 ms | 295 ms | 985 ms |
| Browserbase | 1,669 ms | 1,873 ms | 188 ms | 517 ms | 792 ms | 179 ms | 885 ms |
| Baton | 2,084 ms | 2,191 ms | 245 ms | 271 ms | 1,358 ms | 219 ms | 735 ms |
| Hyperbrowser | 3,666 ms | 5,338 ms | 1,732 ms | 348 ms | 378 ms | 1,200 ms | 3,279 ms |
| AnchorBrowser | 7,919 ms | 11,561 ms | 3,797 ms | 184 ms | 1,260 ms | 2,761 ms | 6,742 ms |
- Without the page load, a Baton session cost 735 ms: second-fastest of the six.
- Loading google.com is where Baton loses: 1,358 ms against 378–1,260 ms. The network is not the cause (first byte after about 75 ms). A browser restored from a snapshot runs cold: its first load of a heavy page takes about 1.3 s, and about 350 ms by the third load. We think nested virtualization makes every first touch of a memory page slow; we have not run on bare metal.
- Different days, instances and run lengths (Steel: 5,000 runs; ours: 300); read it as the same method, not the same hour. Raw runs: browserbench-baton.jsonl.
Moving, copying, sleeping
What the features cost the browser that has them: how long it is paused, and how long a client waits. Live migration under a running client, 0 disconnects.
Live migration
| p50 | p95 | p99 | max | n | |
|---|---|---|---|---|---|
| browser paused | 223 ms | 242 ms | 242 ms | 242 ms | 20 |
| unreachable | 225 ms | 244 ms | 244 ms | 244 ms | 20 |
| client’s longest wait | 595 ms | 669 ms | 669 ms | 669 ms | 20 |
| client round trip, normally | 2 ms | 2 ms | 3 ms | 3 ms | 199 |
| memory still arriving after | 802 ms | 880 ms | 880 ms | 880 ms | 20 |
Fork a running browser
| p50 | p95 | p99 | max | n | |
|---|---|---|---|---|---|
| parent paused, 1 copy | 201 ms | 205 ms | 205 ms | 205 ms | 5 |
| 1 copy ready | 1,448 ms | 1,531 ms | 1,531 ms | 1,531 ms | 5 |
| parent paused, 3 copies | 200 ms | 201 ms | 201 ms | 201 ms | 5 |
| 3 copies ready | 2,607 ms | 2,692 ms | 2,692 ms | 2,692 ms | 5 |
Pause to storage, resume anywhere
| p50 | p95 | p99 | max | n | |
|---|---|---|---|---|---|
| pause: request → stored | 2,120 ms | 2,265 ms | 2,265 ms | 2,265 ms | 10 |
| pause: browser stopped | 409 ms | 416 ms | 416 ms | 416 ms | 10 |
| pause: upload to S3 | 1,704 ms | 1,900 ms | 1,900 ms | 1,900 ms | 10 |
| resume: request → answering | 1,217 ms | 1,297 ms | 1,297 ms | 1,297 ms | 10 |
| resume: download from S3 | 1,016 ms | 1,030 ms | 1,030 ms | 1,030 ms | 10 |
Start from a saved profile
| p50 | p95 | p99 | max | n | |
|---|---|---|---|---|---|
| first session (fetches it) | 1,602 ms | 1,602 ms | 1,602 ms | 1,602 ms | 1 |
| next sessions | 747 ms | 959 ms | 959 ms | 959 ms | 7 |
Starting a browser
From the orchestrator’s side, and from a customer’s: through the internet, Vercel and CloudFront.
Fresh start
| p50 | p95 | p99 | max | n | |
|---|---|---|---|---|---|
| create | 222 ms | 273 ms | 294 ms | 294 ms | 100 |
Public path, as a customer
| p50 | p95 | p99 | max | n | |
|---|---|---|---|---|---|
| create (API) | 64 ms | 290 ms | 1,181 ms | 1,181 ms | 40 |
| connect (CDP) | 93 ms | 117 ms | 298 ms | 298 ms | 40 |
| first page | 265 ms | 339 ms | 527 ms | 527 ms | 40 |
| total | 433 ms | 630 ms | 2,005 ms | 2,005 ms | 40 |
What a browser costs
Memory a browser really costs on its node: private is its own, PSS adds its share of pages shared with others (3 fresh browsers on one node; then one of them forked into 3 copies on that node). Each browser has 1,024 MiB of guest memory.
191 MiB
a fresh browser’s own memory
312 MiB with its share of the snapshot
105 MiB
a fork copy’s own memory
212 MiB with its share
727 MiB
fork image, shared by every copy on a machine
Noisy neighbour
| setup | p50 | p95 | p99 | n |
|---|---|---|---|---|
| hog idle | 4.7 ms | 4.8 ms | 5 ms | 44 |
| hog busy, both VMs in one shared cgroup (no per-VM cgroups) | 4.7 ms | 8.3 ms | 17.1 ms | 44 |
| hog busy, own cgroups, equal weight, no cap | 4.7 ms | 14.7 ms | 15.1 ms | 44 |
| hog busy, capped at 1 CPU, 20 ms period (the default) | 4.7 ms | 5.7 ms | 11.3 ms | 67 |
How, and what to keep in mind
- Machines: m8i.large (2 vCPU, 8 GiB) with nested virtualization; gateway t3.small. VMM: Firecracker v1.17.0 under its jailer; cgroup v2 per VM (guest + 256 MiB; capped at 1 CPU once running); 2 pre-started jailed VMMs kept ready per node. Guest: Chrome for Testing 156, 2 vCPU, 1 GiB per browser.
- Nested virtualization (KVM inside an EC2 VM) is slower than the bare-metal hosts sandbox providers usually run on; the numbers here are likely lower there (not measured).
- The client: bench client, t3.small, same VPC. Internal tests talk to the orchestrator and gateway on private addresses; the public test goes through the internet, Vercel and CloudFront, as a customer would.
- Fresh start is the orchestrator’s own time from request to Chrome answering, including placement, the node starting the VM in its jail, and the network route.
- Percentiles are over the samples listed; the raw samples are in the JSON.