임베드 보드

사용자 등록 기반 큐레이션

Opus 5.5 对比 GPT 6.1 Sol 对比 GPT 6.1 Terra 全新的 GPT Terra 速度快得惊人而且非常便宜!我能预见到这个模型会被应用到许多不同的领域中。 GPT Sol 耗时:30分32秒 成本:$0.96 Opus 5.5 耗时:52分36秒 成本:$15.36 GPT Terra 耗时:3分25秒 成本:$0.22

0 0 0 0

评论 76

  • @itsfazley 回复 ♥ 4

    @rubenssoto_ai Speed and cost are impressive, but what does the quality gap look like? If it's 10x faster but requires manual cleanup, the savings disappear pretty quickly on complex tasks.

    原文
    • @rubenssoto_ai 楼中楼 ♥ 1

      @itsfazley 3D design is a very specific task, with these speed, imagine this model in tasks like computer usage

      原文
    • @rubenssoto_ai 楼中楼 ♥ 1

      @web3devop For sure man, but gpt terra is not supposed to be closer to opus

      原文
  • @jshai0 回复 ♥ 1

    @rubenssoto_ai Speed and cost are fine, what about the results/ outcomes?

    原文
    • @rubenssoto_ai 楼中楼 ♥ 1

      @jshai0 GPT terra is not supposed to be close to these models in quality, but imagine this speed in computer usage

      原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @neerajjj6785 Opus is the best but if you need something fast, terra is the correct one

      原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @KrishnaBuildsAI hahaha I think you need to use Terra on the correct task

      原文
  • @knowixbuilds 回复 ♥ 4

    @rubenssoto_ai Terra is not much of an upgrade from Sol, but its cost is one for the books

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @knowixbuilds I think Terra is the correct model if you need something fast and cheap

      原文
  • @karanjagtiani04 回复 ♥ 1

    @rubenssoto_ai Did all three land the same result? 3 minutes next to 52 is the gap I'd want to diff before comparing the bills.

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @yohakujpn You need to analyze which model is the best for each task...terra can be used for tasks you need something really fast

      原文
  • @VanillaCache 回复 ♥ 1

    @rubenssoto_ai really impressed with the speed and cost with tera here 👌

    原文
  • @YourAKShaw 回复 ♥ 0

    @rubenssoto_ai $0.22 and 3m 25s versus $15.36 and 52m 36s is a massive difference. But the outputs aren't equivalent, which makes this a quality-adjusted cost question. Could you run all three against identical acceptance criteria? That's where Terra's economics get interesting.

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @YourAKShaw Could you guide me how to do that? I have never tested like this

      原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @Eliopyr For sure, Opus quality is much better, Terra helps when you need something really fast

      原文
  • @SahilPanhotra 回复 ♥ 1

    @rubenssoto_ai quality isn't that good but good for quick conversations or small fixes

    原文
    • @rubenssoto_ai 楼中楼 ♥ 2

      @IndependentEco No doubts man, but like you mentioned on your posts terra is pretty fast...and Im going to try some loops

      原文
    • @rubenssoto_ai 楼中楼 ♥ 1

      @LeoBuilds_ For sure mate, Terra has another goal, to be fast and cheap

      原文
  • @BroadsideCode 回复 ♥ 1

    I dont really see how they can tout a faster smaller model as something valuable if every single use case is different for every single user. Like. ok its good and pressing buttons, but how many, how complex and how many times does it need to be fixed after. I dont get the cheap models.

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @BroadsideCode This model is supposed to be fast and cheap, so it's expected that it won't match the quality of Sol or Opus, especially in a single run. From what I've read, it's really good at computer use, and if you work through multiple iterations, you can get much better results.

      原文
  • @pelolisu 回复 ♥ 1

    @rubenssoto_ai Terra is not cheap, 30min on Terra would be ~$2.2. Sol 6.1 in this case is better.

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @pelolisu You should not calculate like this man, terra is much faster

      原文
  • @KylePomykala 回复 ♥ 2

    @rubenssoto_ai I'm doing some similar tests for creative arena - here is what I got so far

    原文
    • @rubenssoto_ai 楼中楼 ♥ 1

      @KylePomykala Wonderful results mate, Im testing Terra with more iterations because you can ask to improve the output since the model is fast and cheap

      原文
  • @aiZgred 回复 ♥ 1

    @rubenssoto_ai Noo u can not say that! It’s better to join a complainers club and complain on every new model released before even testing it 😅

    原文
    • @rubenssoto_ai 楼中楼 ♥ 1

      @aiZgred Oh, sorry, man. I forgot I wasn't supposed to say anything good about the model hahaha.

      原文
  • @Noah_Twenty 回复 ♥ 0

    @rubenssoto_ai Heyy, Haiku would’ve been very relevant to add in this test

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @Noah_Twenty It is a good model to compare, Im going to do that thanks

      原文
  • @AbdoKerdawy 回复 ♥ 2

    @rubenssoto_ai Terra isnt the best but the cost difference IS INSANE

    原文
  • @Ionkosm 回复 ♥ 1

    @rubenssoto_ai Terra is fast but the result... not so much :D At least it was super quick. I think with a few iterations it could scratch the quality of the other two and still being faster

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @Ionkosm For sure man, quality opus is Unbitable, but the comparison was more to show the speed and cost.

      原文
  • @Worldi123 回复 ♥ 1

    @rubenssoto_ai Terra at $0.22 and 3.5 minutes vs Opus 5.5 at $15.36 and 52 minutes is a huge gap. Did the outputs hold up in quality, or did Terra need more retries on the harder tasks?

    原文
  • @magica_ai 回复 ♥ 1

    @rubenssoto_ai Same task, 70x cost gap. Magica routes work to the right model automatically, so your credits go on results and not on picking the wrong tool.

    原文
  • @ceo_multiai 回复 ♥ 2

    Lol, well Opus 5.5 is the leader as always. Its detail is strong. But I got the same result through http://multiai.store, it came out to $2.50 on Opus 5.5, but I feel like nobody's gonna beat that. But if Fable 5.5 comes out, then it'll be interesting to compare these two models

    原文
  • @GoooSAI 回复 ♥ 1

    @rubenssoto_ai I like this)) You can feel the future in it.

    原文
    • @rubenssoto_ai 楼中楼 ♥ 0

      @Smart04261410 The quality for sure, but cost and speed terra is the winner

      原文
스크롤하면 더 불러옵니다