Something is wrong at OpenAI. Tibo and Sam said GPT-6.1 Sol would be faster today. Instead it appears to be running slower. Sonnet 5.5 and Opus 5.5 are much faster than Sol 6.1 and Grok 4.7 Check out the below comparison, same prompt on high reasoning with a 20... minute deadline.show more

Don't worry

Haha this is a great meme

That is clearly not the model 6.1 Sol causing the issue though. 6.1 Sol is so much faster in the benchmarks so this is purely a compute issue. (Likely related to Dots getting free Astra usage is my guess)

Yes dots is the main priority for OpenAI now which is a shame. I had incorporated sol into my daily workflows but I now need to reassess

Speed doesn't really matter as long as things are done correctly; that's what counts. While I haven't tried Sonnet 5.5 yet, Grok 4.7 can be pretty fast, but its output isn't always stable. It doesn't always deliver what I want or expect, meaning its instruction-following can sometimes be on the lower side. Another issue is that fixing one thing can sometimes break another, which is pretty frustrating.

If you want speed and quality opus 5.5 is the model for you

Wow I love it! But yeah Sol is dying on us. :(

Thanks man I’m having fun putting the models head to head

Was the prompt including plan or all went freestyle?

There was no plan I don’t think opus or sonnet need a plan anymore. Maybe sol still needs a plan

Lol they blsme dots. And sell super fast !! The sucl vet out!! Ru y

Dot is the problem for sure. Just put all the compute into sol

Tibo is full of shit.

Why is he lying to us?

Great comparison Eco!!

I appreciate you bro!

Genuinely bro, I also feel Luna is slow

Why are they coming for my Luna 😭

It's better now

I will test out Sol today and see if the problem is resolved

your 20 minute sprint is the fact check on "should be much better now." that's the new normal: official claims get tested by users in real time.

I don’t trust any lab anymore. I’m going to test everything they tell us

Interesting, will be keeping an eye on this.

Let me know if you see an increase in speed

@grok could you explain why

Do you use Dot? I’m trying it tonight it’s fast

Of course dot is fast where do you think my Sol compute went 😂

😂😂😂😂 true. I get the play now. Cut the regular usage, make you use dot who doesn’t work as hard as codex. It’s pretty good though.

I think dot will be a smart business move if they can keep market share away from muse and grok bot

I can see that, cloud computing, file storage, coordination. It’s useful. Def feels like enterprise is the end goal

I think you are right with enterprise being the goal. Watching dev day felt like OpenAI role playing as Google or Apple

thats pretty cool comparison you make this yourself?

I did, if you have any feedback let me know. I’m trying to get better at video

Sol 6.1 is really good but dam, it’s really really slow , i took the entire day to handle a only few tasks

Yeah it is really unfortunate what OpenAI is currently doing with sol

OpenAi get weird day by day. Especially using Sol 6.1 be like having more coffee break

OpenAI was doing so good for so long. I don’t know what happened

OpenAI likes to actively BS their customers. Kind of their thing

Why lie when you can just say nothing lol

Probably to keep subscribers on the hook. Degrade the models mid month, then "fix" them just before renewals so people think things are improving. Toss them some usage. Same cycle, different month. Rinse, repeat. 💀 Difference this time is they degraded from the get go.

It’s such a shame. I used to feel like I could trust Tibo and OpenAI. Now I just feel like they are out to make a buck

Yeah, dev day was just a rug pull and lame copycat products.

Dev day could have been so good if they announced Sol 6.1 and pointed all compute at it

Agree 💯

True. I've been using Codex for months now and was really frustrated with it being slow and expensive. Switched back to Claude and was thrilled how fast it is, and how well it works.

@thsottiaux

@thsottiaux I believe Tibo is working to fix the speed

It's still very slow. 😮‍💨

I'm going to cancel chatGPT.

They are, right? Thought I was going crazy. I feel like 6.1 is not bad, but it's nowhere near astra.

Astra is the better model in my experience

Scamgpt

Only person who isn’t a bought and paid shill on here.

sol is still at the devday afterparty.

🤣🤣🤣

Very nice video 👏

Yeah Opus 5.5 created an android app for me from scratch in 35 minutes

Opus 5.5 is basically magic at apps

It didnt even use an emulator. Just raw dogged an APK file and one shotted it.

Imagine what Fable 5.5 is going to be capable of

I feel sorry for anyone that has an iPhone. They'll never be able to use free custom apps.

interesting can you share the full brief? and opus 5.5 would be better compared to astra no?

Yes Opus 5.5 is the best model in the world. What other data are you looking for?

and if 5.5 is the best why not compare it to astra as well as 6.1. you compared the two anthropic models but only one openai (the lesser one)

I could work Astra into here for sure

slower on the same prompt. loading screen won today.

I put urgency into the prompt which you can tell Opus and Sonnet understood. I wonder if Sol was just confused. Either way Sol is so slow

yeah, let sol have a clean day for once. crypto has already scheduled enough drama.

What is that cost per completion looking like? Speed definitely matters, but it's not the end all if you run out of usage.

Usage is so good on subscriptions for all of these models. I think Sol might be the cheapest

I agree Sol is very inexpensive right now.

Cool bro... respect 🙏

I appreciate you bro

I think if he had just said nothing it would be faster. But tweeting out 'it's faster now' probably added 5x usage and slowed it down again :D.

Lmao you are probably right

It's runnning way faster

You think sol is running faster?