Densa Labs

Densa Labs is an independent software lab building developer tools, experimental software, and research projects. By @rllingho.

[rllingho @rllingho]

Haiku 5.5 now on Benchmark Registry!

CursorBench 4.0 (max): 48.4%
Terminal-Bench 4.0 (max): 39.2%
HLE Full (with tools, max): 57.4%

https://t.co/4QGB0ORfp1

[Densa Labs @DensaLabs]

We have just added 1,872 new results to Benchmark Registry!

Most models now have 10+ results and the Compare feature now has more results to compare from.

https://t.co/kd8b36f0iC

[Densa Labs @DensaLabs]

We have just added 1,872 new results to Benchmark Registry!

Most models now have 10+ results and the Compare feature now has more results to compare from.

https://t.co/kd8b36f0iC

[rllingho @rllingho]

Live:
https://t.co/mVvAMRV9YT

[rllingho @rllingho]

We just added 1,872 new results to Benchmark Registry!

The compare models feature now has more results to compare models to!

See the difference:

[rllingho @rllingho]

Benchmark results are currently scattered across blogs, model cards, reports, etc.

Benchmark Registry puts them in one place.

[rllingho @rllingho]

Benchmark Registry: One place for AI benchmark results.

95 models, 84 benchmarks, 942 results.

https://t.co/mVvAMRV9YT

[Densa Labs @DensaLabs]

Available today:

https://t.co/FPe135La9y

[Densa Labs @DensaLabs]

Install on a Mac in one line:

curl -fsSL https://t.co/6SBpCpJgaG | sh

Homebrew, Windows, and Linux too:
https://t.co/FPe135La9y

[rllingho @rllingho]

We’ve just added 290 new results to Benchmark Registry!

Models that used to have less than 3 results now have all results we can find from their system cards/blogs.

https://t.co/mVvAMRV9YT

[Densa Labs @DensaLabs]

Gemini 4 Argon now on Benchmark Registry!

AutomationBench: 51.3% (SOTA)
DeepSWE 1.1: 77.9% (SOTA)
Terminal-Bench 4.0: 57.4%

https://t.co/jR6bxOZEc9

[Densa Labs @DensaLabs]

Gemini 4 Argon now on Benchmark Registry!

AutomationBench: 51.3% (SOTA)
DeepSWE 1.1: 77.9% (SOTA)
Terminal-Bench 4.0: 57.4%

https://t.co/jR6bxOZEc9

[Densa Labs @DensaLabs]

We’ve added 290 new results across 39 models to Benchmark Registry!

All data is sourced from official model developer system cards or blogs.

Check it out:
https://t.co/kd8b36f0iC

[Densa Labs @DensaLabs]

We’ve added 290 new results across 39 models to Benchmark Registry!

All data is sourced from official model developer system cards or blogs.

Check it out:
https://t.co/kd8b36f0iC

[rllingho @rllingho]

Just launched https://t.co/mVvAMRUC9l on Product Hunt!

https://t.co/88KLFKhD9A

Benchmark Registry: One place for AI benchmark results.

Would greatly appreciate support and feedback :)

[Densa Labs @DensaLabs]

Now in Benchmark Registry:

- Compare Models
Compare models side by side to see the difference in their benchmark results

- Explore Benchmarks table
- Latest Additions table
We’ve replaced the old Recent tables to make Benchmark Registry more useful

https://t.co/kd8b36f0iC

[Densa Labs @DensaLabs]

Now in Benchmark Registry:

- Compare Models
Compare models side by side to see the difference in their benchmark results

- Explore Benchmarks table
- Latest Additions table
We’ve replaced the old Recent tables to make Benchmark Registry more useful

https://t.co/kd8b36f0iC