Show HN: Abliterated LLM provider for cyber tasks

Posted by jurisdb on Monday October 05, 2026 @08:36AM
from the show-hn dept.

https://refuseless.com

Current lineup includes GLM 5.3 and its Flash variant. MiMo V2.6 and Qwen Flash Next will be added in the near future.

( 6 comments | 4 points | on HN )

Re:Show HN: Abliterated LLM provider for cyber tasks by freakynit on Monday October 05, 2026 @09:04AM (#49964319)

Signup page is throwing 500 error after google login/signup...

[ Parent | on HN ]

Re:Show HN: Abliterated LLM provider for cyber tasks by json-all-in-one on Monday October 05, 2026 @09:42AM (#49964796)

Same here, no way to test it

[ Parent | on HN ]

Re:Show HN: Abliterated LLM provider for cyber tasks by jurisdb on Monday October 05, 2026 @09:55AM (#49964956)

Apologies, we were deploying some changes. Should work now.

[ Parent | on HN ]

Re:Show HN: Abliterated LLM provider for cyber tasks by mdebeer on Tuesday October 06, 2026 @11:12AM (#49979714)

Are you abliterating these yourself or using community versions?

[ Parent | on HN ]

Re:Show HN: Abliterated LLM provider for cyber tasks by jurisdb on Wednesday October 07, 2026 @09:07AM (#49992243)

We use very popular community versions

[ Parent | on HN ]

Re:Show HN: Abliterated LLM provider for cyber tasks by ElementMerc on Friday October 09, 2026 @11:34AM (#50022041)

Any benchmarks or checks to ensure the capability wasn't affected too badly after abliteration? If not, you could check that out using Senbonzakura (https://github.com/elementmerc/senbonzakura). It's free

[ Parent | on HN ]