DeepSeek V4.1 Flash 1017 MAX vs Openai Gpt 6.1 Sol Extra High DeepSeek Input Tokens: 1.970.140 Output Tokens: 243.599 Time: 8 hours and 6 minutes Cost: $0.96 Gpt Sol Input Tokens: 96.063 Output Tokens: 54.730 Time: 20 minutes and 37 seconds Cost: $1,29 Which one do you think is... the best?show more

8 hours vs 20 minutes is diabolical

I was using opencode free, maybe that was a provider problem

I am gonna say DeepSeek better, but GPT 6.1 Sol still very impressive!

Both are really good, but deepseek is more cozy :)

love that you posted tokens and time, not just the price. great test. 8 hours vs 20 minutes for a 33 cent difference is the real tradeoff here.

But don't take the time too seriously, could be a problem in the provider, im not sure

I have to say DeepSeek really nailed it

I agree, really good quality

8 hours.....cheap, but time is money

I think it was some problem with the ai provider I was using opencode free

deepseek did more

I liked deepseek the most too

DeepSeek wins on cost, GPT on speed.

For sure man, but I think there was something wrong in the deepseek provider

Deepseek won

I agree mate

I like the DeepSeek's mood a lot more

I also liked, much more cozy :)

Deepseek looks more realistic on the water fall. Also, the plants moving a bit like they are blown by the wind.

I liked deepseek the most because it seems more cozy :)

The 8-hour vs 20-minute gap is what I keep staring at. I ran a similar head-to-head on my video renders last week and the slow one won on detail, so I will take cheap and patient over fast.

I think the provider was facing some problem, because this amount of time is not normal

That was my first thought too. I once reran a benchmark three times before I realized the provider queue was the real bottleneck, not the model. Numbers that skewed usually say more about that day's infra than the model itself.

Money loves speed

I agree, I need to try more, maybe it was a problem on the ai provider

Blended that is about $0.43 per million tokens for DeepSeek vs $8.55 for GPT, but 20x the input tokens leaves the per-run bill at $0.96 vs $1.29. Price per token is not price per task

For sure, price per task is much more important

Sol wins this one for sure

I liked deepseek cozy environment haha

Deepseek 8 hours seriously man? It created a whole village though!

I really dont know what has happened, did you use the MAX mode? I started 6 am and just finished now

Haven’t tested MAX yet! 8 hours is crazy, but for under $1 to get that level of scale, I definitely need to try it.

I need to test if there is a speed different between Max and other models...but yeah could be something in the provider, opencode free, is free haha

Could be a provider issue!

Paying 33 cents to get almost 8 hours of my life back. Sol.

To be fair, maybe was a problem in the provider, I was using Opencode free

8 hours 🫠🫠

Maybe the problem was the provider, I was using opencode free

Deepseek has a way better feeling into it. Something like animal crossing vibes.

Deepseek much more cozy :)

The details come from editing with the llm? Or better algo to render?

Everything was created by the llm using three.js, I didnt use any external tool

Why did DS take so long? It doesn’t seem like a ton of toks

I think it was a problem in the provider but Im not sure

so awesomeee!

yes, I loved it

so dreamy

关键还是任务和输出质量得对齐，不然光比 token 数看不出谁更省。

Perfect and I think this time deepseek was cheaper and with better quality

Now DeepSeek is sick! 🔥🤯

Deepseek is a great model for sure

Find the DeepSeek more finished

I agree, it seems more cozy, more details

where the water coming from ? 😭😂 nice design btw

this is magic my friend haha...thank you

8 hrs 🙂

I think it was some problem with the ai provider I was using opencode free

DeepSeek consistently renders thin "wispy" water and fire by default on runs I've seen. Probably matches some specific cartoony aesthetic taste, but it looks unrealistic.

I get what you are saying, there is a video I posted yesterday with realistic water created with deepseek, it was not easy, multiple iterations

I like DeepSeek more

Very cozy, right?

Yh

GPT Sol is the winner here, faster and cleaner

Something strange with the time deepseek took to finish, need to take a closer look

Deepseek looks better until you look closely. The fence etc kills it

oh man, you have a very good eye, the fence doesnt make any sense haha

The cabin logs go through the windows 🤣

damn this looks really good for deepseek and the cost is big factor here.

The deepseek output was better, I agree

deepseek looks unbeatable on cost and throughput here, but the real winner depends on whether both delivered the same quality on the actual task

I liked deepseek the most, it seems more cozy, it took 8 hours but I think it was a problem in the provider

Token efficiency is becoming a real competitive edge, the best model is the one that delivers quality without burning time and budget

Deepseek is token hungry, but the time maybe it was the provider slowness Im not sure

I can easily tell which one is DeepSeek just from the output

How man?

The output quality

They are both sooo good

Year I liked both, deepseek is more cozy

I agree