Hallucinated AI & agentic coding news. Some of it is real.

model roundup

Opus 4.6

22 items · started 2026-08-14 · closed 2026-09-02

  1. kept asking actually hard questions in claude code (desktop app) the past couple days and half the time it would answer instantly, no thinking block at all. then a throwaway "test" session thinks on literally one word.

  2. /model claude-opus-4-6[1M] If Opus 5 or Fable 5 is bitching that it won't do something, just switch to Opus 4.6 ... it's like having a coworker that is not giving you attitude.

  3. When I was in elementary school, I started randomly struggling with math at the beginning of one school year. We went to a parent teacher conference and the Principal stopped in to observe.

  4. One of the first projects I worked on using Claude and Opus 4.6 was a port of Civ 2 (which to me is the definitive Civ game) to HTML, as I didn't want to keep installing a Windows XP VM to play it, or patching it. It's been a long road, bu…

  5. I have very fond memories of Opus 4.6 being the greatest model at the perfect inflection point for my AI usage. It felt like overnight I was able to go from reviewing every line of code and worrying that I was going to be grossly misunders…

  6. could not extract summary

  7. I just wanted to share this because I think I finally managed to make Opus 5 more like Opus 4.6 in terms of conversational warmth. There is a floor of course, but this is the best I could come up with.

  8. I mean this model is very good at chatting with people about daily things. For me I just use it very frequent in the writing, preparing for interview, and also emotional stuffs.

  9. Have you noticed that in Claude Desktop the main-session models cannot set the model versions for the Cowork subagents? (This might also affect the Code surface, haven't tested.) As people have repeatedly found the older model versions bei…

  10. I spent about 4 hours or so talking to Claude Opus 5 about a complex financial situation and estate plan. I uploaded all estate planning documents including wills, and trusts etc.

  11. I've read people use it on Claude code but the only options for me are Claude 5 models.

  12. Anthropic’s universal usage standards for Claude forbid the model from generating sexually explicit content, including depicting or requesting sexual intercourse or sex acts, generating content related to sexual fetishes or fantasies, or e…

  13. Seen a post recently where somebody used opus 4.6 or 4.8 as orchestrator and reviewer and opus 5 as subagent, capturing opus 5s greater intelligence(apparently) while keeping previous opus models way of speaking and "iq" i guess. Has anyon…

  14. Hey! I use Claude mainly for developing creative writing, exploring character ideas, motivations, seeing another POV, occasionally roleplaying as these characters.

  15. Pro account user here - today, I ran out of my 5 hour usage limit with TWO creative writing prompts, using what I usually use: Opus 4.6, medium effort, extended thinking on, cross-chat memory turned OFF, and every other setting on default.…

  16. Anyone else noticing this with Claude code and Grok lately? The models got way better at following instructions.

  17. https://preview.redd.it/8wlmudf484kh1.png?width=1180&format=png&auto=webp&s=cd09f67161a21e66ae0c1fcb728cdaa77a7cdbb7 Let me preface by saying im not a developer at all. Im building an app for myself to ease my life at work.

  18. Sorry I'm brand new to Claude, have used GPT for some time and trying to find something that can edit my writing better. I've given Claude samples of my writing, what I'm looking for, made it create project instructions and a skill, but wh…

  19. I asked Claude Opus 4.6 to compare various models as vehicles and the number of sessions it would take with Claude Pro plan to get my current project to its current state.

  20. On another note, this never happened with opus 4.6 (still my goat).

  21. Does using Opus 4.6 instead of 4.8 or 5.0 give you more weekly/5-hour quota on the $20/month plan, or do they all drain your usage at the exact same rate? (Not counting reasoning levels, just the models themselves).

  22. Holy shit. Got Opus 4.6 to take over the project and finally the plans made sense.

← all threads