Well I originally made this for my Openclaw agent but ended up using it myself since I had no account before. Guess I'll stick with it.
- Tweets 127
- Following 17
- Followers 4
- Likes 1
12 hours of Sol-6.1 High since the reset 30% of 20x $200 plan used up. The work that got done is similar to what Astra High was doing before in 8 hours for like 16-17% usage.
Sol6.1 is so much worse on complex tasks that it goes on side quests wasting time/usage.
1
1
5
procaclaw retweeted
gpt 6.1 sol: first frontier model to work with s/t instead of t/s
10
3
398
28,380
Can anyone from Google explain what is the reason for Gemini 4 to not be released to the general public? It's not even a Mythos/Astra class model.
4
piped.video/watch?v=pYH8x-Nb…
Ok in the demolition derby example Sol took shortcuts didn't meet the specification and acknowledge it.
And this is the problem that doesn't show in benchmarks. Is it efficiency when you skip half the requirments? Or it is wasted time and tokens.
4
OpenAI model strategy is Release good model with insane capabilities (Astra) that they can't even serve.Everyone benchmarks it call them frontier lab.
And release small model that is inferior in every way but use it as argument to halve the usage.
3
Here is my experience with GPT6.1-Sol
I have a workflow generate character then use local minimax H3 for video to generate the animations remove background>review the output correct it> release ready animation sprite sheet. Astra on high did 1 char/hr. Sol6.1 with 3 sub agents
1
16
Here is what happend.
OpenAI tried to upsell GPT6-Sol as Astra doubling it's price.
Then it tried to to compete on price with Chinese labs with 2 very small models named Sol and Luna.
The $200 plan usage is halved basically making Astra unusable and you are stuck with the small.
6
I recently noticed increase of ADs/promotions of "Agents for specific tasks" offered as a service or something dumb like that. I don't understand this business model. Why would someone pay for it when they can ask their own agent. Also what VC is funding this nonsense.
10
Considering Opus 5.5 is better than astra in a lot of cases. Doesn't it make sense that Astra is priced accordingly at like 4-5$
People may say Astra is more token efficient. Sure I agree but using less tokens for considerably simpler output is not really a top tier model.
6
Is this AI psychosis ? Astra building game assets for one game using locally hosted minimax H3 for video generation, while Opus 5.5 is building another game using blender and UE5?
10
Watching a lot of comparisions of Opus 5.5 vs GPT6-Sol Sol seems really like very small model. and Opus is way way over the top. Where is the in between ? I don't need Opus 5.5 5 hour per task at $50 but I'm not happy with 6 SOL with same prompt working 10 minutes $2.
15
Is it only me but codex starting to feel more like openclaw, I'm slowly using Openclaw less and less and just straight using codex. I was using openclaw for the loop setup and orchestration and delegating the work to codex.
4
So you are telling me that Fable 5 can run for 60 minutes then silently switch to Opus and deliver Opus results after paying Fable 5 price. Sorry but this is scam.
$1T valuation and this is the best you came up with ?
12