Skip to content
DurkBench

Grok Build

Grok 4.5

Inside Grok Build we ask for grok-4.5 every hour. Everything below comes from those runs on our lab machine.

Grok Buildgrok-4.5Think setting highLast round just nowOther models

Write speed, last 24 hours

42.2wall tok/s

Tokens per second across the whole command — you press enter, the app starts, thinks, writes, finishes. Higher is better. This is the middle value of 33 clean rounds out of 33 tried in the last 24 hours.

Time to first word
3.98s
How long before the first visible word shows up. Lower is better.
Middle wall time
6.68s
How long the whole command took, start to finish, including the app booting. Lower is faster.
Last ping
The newest tiny “reply ok” check. It measures wake-up time, not writing speed.

What stands out

Four facts pulled straight from the saved runs.

Board position
#2 of 10
Just ahead: Haiku 4.5 at 59.2 wall tok/s. Just behind: Grok 4.6 at 41.8 wall tok/s.
Rounds that worked
33/33
Clean rounds out of rounds tried, in the last 24 hours.
Served as
grok-4.5-build
The model name the app reported back. If it differs from the pin, we say so.
Round-to-round swing
4.0×
How much the fastest round beat the slowest one. A big swing means a noisy day.

Recent rounds

Newest first. Fails stay in the list with their reason.

WhenResultWall tok/s
just nowokplugins=2:codex,cloudflare,hooks=1; no tool calls46.0
1h agookplugins=2:codex,cloudflare,hooks=1; no tool calls34.0
3h agookplugins=2:codex,cloudflare,hooks=1; no tool calls42.4
4h agookplugins=2:codex,cloudflare,hooks=1; no tool calls35.6
5h agookplugins=2:codex,cloudflare,hooks=1; no tool calls43.5
6h agookplugins=2:codex,cloudflare,hooks=1; no tool calls27.5
7h agookplugins=2:codex,cloudflare,hooks=1; no tool calls37.1
8h agookplugins=2:codex,cloudflare,hooks=1; no tool calls20.1

All rounds are in Logs.

Read next

Compare with GPT-5.5

Its closest rival in Codex — same window, same units, with the leader marked on every row.

Compare with Grok 4.6

The pin next to it in Grok Build. See what the tier jump buys inside one app.

More models

How to read this

Every number here comes from our lab machine running the paid app, one try per hourly round, on test kit v6. Our measurements, not the vendor’s.