Compare
0
×
Feedback
Send feedback
Share product ideas, UI feedback, or workflow requests.
×
Your feedback goes straight to the team.
Send
InferHub
Open vLLM and SGLang inference benchmarks, contributed by the community.
Server Launch Configs →
Checking sign-in…
?
Sign in
Create account
Account
▾
Edit display name
Sign out
Official
Team
My Workspace
loading...
+ Add a benchmark result
Add a benchmark result
Paste or upload your benchmark output.
Use the CLI skill instead →
Faster: use the inferhub Claude Code skill
1
Install once:
git clone <repo> && ./install.sh
2
Submit logs:
Use inferhub skill for ./my_benchmark.log
3
Or run+submit:
Use inferhub skill
→ choose "Run new benchmark"
The skill handles hardware detection, server launch, and submission automatically.
▴
Server launch command
Optional, you can modify this via Edit.
Benchmark launch command
Optional, you can modify this via Edit.
Benchmark result
(REQUIRED)
↑ upload file
Hardware info
RECOMMENDED
Run this one-liner on your GPU machine to collect hardware info automatically.
Server
Copy
Waiting for hardware data...
Data quality warnings
Why this run? (optional)
Add Run
Browse and compare runs
Throughput
Mean TTFT
Mean TPOT
Refresh
Best throughput (tok/s)
-
tok/s
Lowest mean TTFT
-
ms
Lowest mean TPOT
-
ms
Benchmark library
?
-
means data was missing or could not be parsed.
?
Folded
groups similar deployment variants into a single row. Click
Expanded
to see every run individually.
Export
▾
CSV
JSON
Markdown
Timeline
Expanded
Folded
Showing duplicate group (
0
records with the same benchmark log)
Back to full table
Throughput (tok/s)
Mean TTFT (ms)
Mean TPOT (ms)
Engine
Source
HF model
GPUs
I/O len
Throughput (tok/s)
Mean TTFT (ms)
Mean TPOT (ms)
Actions
Details
All
vLLM
SGLang
All
Official
Custom
Sign in
×
Sign in
Create account
Continue with GitHub
First time may take a few seconds.
or use email
Email
Password
Display name
(optional, shown next to your runs)
Sign in
At least 8 characters.
Account settings
×
Display name
Save
Shown next to records you create. Your email stays private.
API tokens
Tokens let the CLI and probe script submit data on your behalf.
Copy this token now — it won't be shown again.
Copy
Create token
Compare benchmark runs
Clear selection
×
Keyboard shortcuts
Browse faster without losing the mouse-friendly path.
×
j
Move to the next run
k
Move to the previous run
/
Jump to search
1
Open Official workspace
2
Open Team workspace
3
Open My Workspace
Esc
Close the open panel or dialog
?
Show or hide this guide