Widda Alatta

[Widda Alatta @WiddaAlatta]

@HermesAgentTips gpt-5.5 is solid, qwen-3.6-plus as fallback

[Widda Alatta @WiddaAlatta]

@DavidOndrej1 Have never come across that before, what model are you using?

[Widda Alatta @WiddaAlatta]

@simpsoka The runtime environment is not clear, even if the default working env is WSL, it is not fully Windows native

[Widda Alatta @WiddaAlatta]

@horstenegger @NousResearch On Discord, more than one instance

[Widda Alatta @WiddaAlatta]

@rileybrown Don't mention channels issue, Discord/Telegram. You spend more time fixing it instead of using it.

[Widda Alatta @WiddaAlatta]

@Ryan_Ravenson ACP?

[Widda Alatta @WiddaAlatta]

@EXM7777 Try Discord, it far beyond awesome.

[Widda Alatta @WiddaAlatta]

@sama I hope we get a mobile app for Codex soon, push notifications and workflows on the go would be an awesome addition to the Codex ecosystem.

[Mustufa Khan @mustufa4socials]

https://t.co/m5leqLx6zj

[Widda Alatta @WiddaAlatta]

@fekdaoui Hermes' architecture is pretty solid, and the Python backend is more predictable than JS, I guess. But the effort around OC is still huge and totally respectable.

[Widda Alatta @WiddaAlatta]

@mr_r0b0t @NousResearch @Teknium The tools they provide are unique, a peace of mind, how did you find the inference, consumption, limits?

[Widda Alatta @WiddaAlatta]

@Abdullah4AI فعلا افضل بمراحل...ميزة عرض الأعمال الجارية بالخلفية لوحدها لا تقدر بثمن...ناهيك عن التعلم التلقائي.

[Widda Alatta @WiddaAlatta]

@Teknium @NousResearch That was fast ⚡️, keep it up.

[Widda Alatta @WiddaAlatta]

@makulas1913 يستهلك توكن اتوقع ايضا من خلال مزودي VLM, Embedding

[Widda Alatta @WiddaAlatta]

@sama Bring it on 🔥, when 5.5?

[Widda Alatta @WiddaAlatta]

@daniel_mac8 @openclaw does anyone test this?

[Widda Alatta @WiddaAlatta]

@steipete Fingers crossed, hopefully the GPT model's performance issues are resolved in the next release!?

[Widda Alatta @WiddaAlatta]

@NousResearch While some may perceive this as a shift towards a closed-source methodology, it is important to clarify that this is not the case. The supplementary cloud-based enhancements significantly augment the capabilities of the Hermes agent.

[Widda Alatta @WiddaAlatta]

@KSimback I have tried gpt-5.4, gpt-5.3-codex, MiniMax and more than one glm model in OC, found glm-5-turbo better at coding, and glm-5.1 is the best for frontend/UI. On Hermes however the gpt models seems a bit better as main model, glm-5 series comes next but not MiniMax.

[Widda Alatta @WiddaAlatta]

@openclaw does this fix gpt-5.4 and gpt-5.3-codex performance issues, these particular model perform with less competence compared to other models.