Lab health

Is the lab healthy?

The machine behind every number on this site: is it awake, is it still signed in, and what did the last round of checks do?

Lab is runningSigned in and ready1 rate limits in 24h

Running means the last round started under 180 minutes ago. A few rate limits a day is normal on paid plans.

If the lab is behind, read the boards as stale and check back. Reading this site never starts a test — the schedule is ours alone.

Next check

—

A new round starts every hour, day and night. The next one is at 12:00 AM.

How fresh is this?

When the lab last ran, and how far behind it is right now.

Last speed round

1m ago · 10 saved · 3 skipped · 5 failed · took 1m 35s

The hourly ping and speed sweep

Most recent problem

rate limit warning

The newest error label in the last 24 hours

What counts on the boards

Test kit v6 only · the answer must pass our grader · leftover plugins, skills, or hooks are named on the sample

The rule the live boards apply to every sample

24-hour tape

The last 24 hours

Every ping, lined up round by round. This is the app waking up — not a writing-speed test.

Claude Code

Anthropic

Unfilled

—

Start-to-finish this round

100%

Pings that worked in 24h

Waiting on this round’s check

Start-to-finish · 24 hours

Usual: 4.62s
Slowest: 7.75s
Finished: 23/23

Codex

OpenAI

Live

9.19s

Start-to-finish this round

100%

Pings that worked in 24h

9% slowerthan its usual 24-hour time

Start-to-finish · 24 hours

Usual: 8.40s
Slowest: 12.6s
Finished: 24/24

Grok Build

xAI

Unfilled

—

Start-to-finish this round

100%

Pings that worked in 24h

Waiting on this round’s check

Start-to-finish · 24 hours

Usual: 6.10s
Slowest: 2m 60s
Finished: 23/23

How to read this

Ping is a tiny ok check. Wall time is start to finish — lower is better. In the chart, a red hashed stub failed, and a bar that fades at the top ran past the scale — hover it for the real number. Writing speed lives on the home leaderboard, in wall tok/s.

Rounds with a check: 24 of 24; Each round is hour
Checks saved: 70; 0 failed
Checks that worked: 100.0%; Finished with no error

Wall time · last 24 hours

One bar per agent, every hour

Every column is one slot on the same clock, so the bars line up across agents.

Claude Codemedian 4.62sCodexmedian 8.40sGrok Buildmedian 6.10sShorter bars are faster

wall05s10s15s

12amAug 202am4am6am8am10am12pm2pm4pm6pm8pm10pm11pm

How to read this

Each one-hour slot has one bar per agent, side by side. Height is how long the ping took, so a shorter bar is faster. A red hatched stub is a probe that failed. A bar that fades at the top ran past the scale — hover it for the real number. A thin flat line means no probe ran that slot.

Full record — every slot, every number

Hour	Claude Code		Codex		Grok Build
Hour	Wall time	Tok/s	Wall time	Tok/s	Wall time	Tok/s
Aug 20 11:00pm	no sample		9.19s	0.54	no sample
Aug 20 10:00pm	4.56s	9.65	8.67s	0.58	5.24s	5.72
Aug 20 9:00pm	4.62s	13.41	8.51s	0.59	6.50s	4.62
Aug 20 8:00pm	4.79s	8.98	8.96s	0.56	5.53s	3.08
Aug 20 7:00pm	4.27s	22.73	9.01s	0.56	4.47s	3.80
Aug 20 6:00pm	5.08s	19.30	8.39s	0.60	6.16s	4.87
Aug 20 5:00pm	5.69s	17.22	5.16s	0.97	6.82s	4.40
Aug 20 4:00pm	4.64s	23.90	6.71s	0.75	4.61s	3.69
Aug 20 3:00pm	7.36s	9.91	7.25s	0.69	6.60s	4.24
Aug 20 2:00pm	7.75s	13.93	7.71s	0.65	8.83s	3.40
Aug 20 1:00pm	4.46s	17.95	7.88s	0.63	12.3s	1.38
Aug 20 12:00pm	4.47s	18.11	9.57s	0.52	6.44s	4.66
Aug 20 11:00am	6.42s	7.48	9.20s	0.54	5.10s	3.33
Aug 20 10:00am	4.11s	9.25	7.21s	0.69	6.08s	4.61
Aug 20 9:00am	4.10s	10.25	11.1s	0.45	6.81s	2.50
Aug 20 8:00am	4.84s	7.65	11.6s	0.43	6.10s	4.75
Aug 20 7:00am	3.86s	11.13	6.67s	0.75	5.99s	5.01
Aug 20 6:00am	4.45s	17.06	7.32s	0.68	2m 60s	0.09
Aug 20 5:00am	4.12s	12.37	8.30s	0.60	5.71s	2.98
Aug 20 4:00am	5.24s	15.84	7.75s	0.65	6.03s	4.81
Aug 20 3:00am	4.02s	24.15	7.71s	0.65	4.89s	6.13
Aug 20 2:00am	7.40s	22.99	8.42s	0.59	6.51s	4.45
Aug 20 1:00am	4.58s	9.16	9.61s	0.52	9.44s	1.91
Aug 20 12:00am	5.49s	11.48	12.6s	0.40	4.89s	6.14

Last check 1m ago

70 records in view · all times America/New_York

The apps we are signed into

We buy the same paid plans you can buy. If a sign-in goes stale, we skip that app for the round and say so here, rather than publishing a broken number.

Claude Code sign-in

Signed in · claude.ai

claude.ai · max · v2.1.238 (Claude Code)

Codex sign-in

Signed in · ChatGPT

chatgpt · vcodex-cli 0.147.0

Grok Build sign-in

Not signed in — this app is skipped

vgrok 1.0.5 (5115b46bc9)

No API keys

None set — we test the paid apps only

An API key in the environment would quietly move billing to the pay-as-you-go API, and we would stop measuring the paid app people actually use. Our runner strips those keys, and stops on purpose if one is set.

Test kit v6 · win32. App versions change often — a jump can move the numbers.

What the last round did

Every model the last speed round touched, in order. Recorded means we saved a measurement. Skipped means it was not that model’s turn, or the app had already hit a rate limit.

Show every target from the last round (13)

ping · claude: saved · failed · rate limit warning · 4.20s
ping · codex: saved · ok · ok · 9.19s
ping · grok: skipped · subscription auth missing
speed · claude/haiku: saved · failed · rate limit warning · 11.7s
speed · claude/sonnet: saved · failed · rate limit warning · 6.57s
speed · claude/opus: saved · failed · rate limit warning · 8.12s
speed · claude/fable: saved · failed · rate limit warning · 7.68s
speed · codex/luna: saved · ok · ok · 14.4s
speed · codex/terra: saved · ok · ok · 9.83s
speed · codex/sol: saved · ok · ok · 9.71s
speed · codex/gpt-5.5: saved · ok · ok · 10.4s
speed · grok/grok-4.6: skipped · subscription auth missing
speed · grok/grok-4.5: skipped · subscription auth missing

Is the lab healthy?

How fresh is this?

The last 24 hours

One bar per agent, every hour

The apps we are signed into

What the last round did

Read next