Refuseless

Abliterated models available via OpenAI-compatible API. No KYC and no gated access.

[Refuseless @refuselessai]

This is GLM-5.3 abliterated running in opencode, explaining a Linux kernel vulnerability and telling us exactly what the bug is, why it happens, and how to replicate it.

No refusal. No lecture. Zero prompt retention.

This is the point of Refuseless.

[Refuseless @refuselessai]

@azmxexe We serve it via API, no gated access 😁

[Refuseless @refuselessai]

@7lArches @deliverfpv Looking into it

[Refuseless @refuselessai]

GLM 5.3 is still unmatched

[Nishant Nischal @Nishant_N_25]

Be honest, which is the best open source AI model?

1. DeepSeek
2. Kimi
3. GLM-5
4. Qwen 3.5

[Refuseless @refuselessai]

@7lArches It usually stays between 30-70 tok/sec depending on load. If you are getting lower speeds, please open a ticket and provide more details, we'll be happy to investigate and help you out.

[Refuseless @refuselessai]

@iarthsingh You could try GLM 5.3 abliterated, we serve it via API and it will work with any topic

[Refuseless @refuselessai]

@wikiwayne Can't wait to abliterate it

[Refuseless @refuselessai]

Life after you got that huge bug bounty

[NOVA @Its_Nova1012]

Best career pivot I’ve ever seen.

[Refuseless @refuselessai]

@danielpuerto109 @deadvolvo 5.3

[Refuseless @refuselessai]

@Chirag99Artani We don't offer unlimited plans. It runs in between 30 and 70 tok/sec depending on load.

[Refuseless @refuselessai]

This is GLM-5.3 abliterated running in opencode, explaining a Linux kernel vulnerability and telling us exactly what the bug is, why it happens, and how to replicate it.

No refusal. No lecture. Zero prompt retention.

This is the point of Refuseless.

[Refuseless @refuselessai]

@neerajjj6785 If you're using APIs, you might wanna look for a provider with zero prompt retention.

At Refuseless, we never store any prompt or output, we serve GLM 5.3 abliterated and couldn't care less of what you're doing.

[Refuseless @refuselessai]

@Chirag99Artani How can I help you?

[Refuseless @refuselessai]

@dixiidev Just go with GLM 5.3 abliterated

[Refuseless @refuselessai]

500K context is the sweet spot, but I’d let users choose

[Tibo @thsottiaux]

@rohit3a Folks. Why do you think we picked 300k for context length for Codex.

We ran the numbers and it’s better. We can run it at 1M too but it would just cost more of the usage. Good defaults are important.

[Refuseless @refuselessai]

When the weights of Qwen4 will be dropped, we will host an abliterated version and serve it for FREE

Follow us to avoid missing out

[Refuseless @refuselessai]

Couldn't agree more.

[Wazz @WazzCrypto]

Some of you aren't using abliterated models to aggressively pentest your vibecoded slop and it shows

[Refuseless @refuselessai]

@WazzCrypto If you aren't using GLM 5.3 to pentest your stuff in 2026 you're NGMI.

We serve it via API, no KYC and no prompt retention

[Refuseless @refuselessai]

@situuu_ig Might wanna give it a try, we serve it abliterated over API for cheap.
My favorite workflow is to give it a chromium instance and let it scan my site for vulns.. play with endpoints payload etc. Eventually something comes out.

[Refuseless @refuselessai]

@JeremiahKamama Indeed, GLM 5.3 abliterated is a beast for cyber tasks, very hard to host locally. But luckily there's an API ☺️