← Back to context

Comment by msh

4 hours ago

I dont understand why there is not a gpt-6 terra?

It wasn't really used enough and it sat in an awkward middle space between luna and sol where either luna high/xhigh or sol med were better cost/perf wise

  • I don't agree. In "none" thinking mode, Terra serves a useful purpose where medium-grade intelligence is needed. Luna doesn't cut it.

    Hardly enough time had passed to develop the data to come to a conclusion. Users can take time to build interest.

I'm scared now, my employer only allows Luna and Terra on 5.6. I really hope they will allow Sol then on GPT 6.

Funny thing is they very recently also set a real limit per-user/month, so why even limit the models because theyre "too expensive".

  • Your employer should reconsider. Sol high is cheaper than Terra max and smarter, when measured per task. ie even if tokens are more expensive Sol can often do a job with fewer tokens.

I read that there are rumors that they're getting rid of that tier. No idea where the rumor came from, though. This lends credence to it, I suppose.

It probably didn't see that much use, as it struggled to find a niche. If you wanted intelligence tasks, Sol was cheap enough and much smarter. If you wanted performance and cost-effectiveness, Luna was significantly better value while being only a little less intelligent.

Terra ended up just being an awkward middle ground that was not particularly suited for any workload.

  • I disagree. After a bit of experimenting, I actually found Terra to be a very good workhorse model on none-to-medium reasoning, and I actually quite prefer its code to Sol's in many cases. It has less of a complexity to over-complicate things. Where Sol would have a sea of try/except and recoveries for situations that are structurally impossible, Terra would just write nice, sequential code.

    Maybe for one-shotting large things Sol is better, but for prod code where I decompose into smaller tasks and read all the code I favored Terra.

    Sol 5.6 was still king for architecture/research in my workflow, though.

  • I have found it worked quite well as the workhorse model in my hermes agent.

  • The users of Terra disagree. Specifically, Terra is useful when medium-grade intelligence is needed in instant ("none" thinking) mode.

    Hardly enough time had passed to develop the data to come to a conclusion. Users can take time to develop an interest.