DeepSeek's New AI Is An Absolute Monster
The new DeepSeek model delivers a massive capability jump in just five weeks through self-feedback learning rather than architectural changes, offering high performance at low task costs.
Highlights
2 clips · 3 min of 6 min-
- DeepSeek's API pricing is significantly lower than competing frontier models.
- Comparing costs strictly by price per million tokens can be misleading due to token efficiency differences.
- DeepSeek can require two to three times more tokens per task, making cost-per-task the more accurate metric.
-
- The new version achieved a massive performance boost with zero architecture changes.
- Rather than relying purely on external training data, the model learns by critiquing its own output.
- Self-feedback allows the AI to iteratively test, catch errors, and fix its own generated code and simulations.
That’s all the highlights.
Questions this video answers
10 · every answer cites the moment it was said01 How do token-based costs compare to actual task execution costs? 2 sources · 1:08
While token pricing looks extremely cheap on paper, different systems consume varying amounts of tokens for the same job. DeepSeek can take two to three times more tokens than Claude Opus for identical tasks, making task-based metrics a truer cost measure.
02 How much does it cost to run DeepSeek via the API? 2 sources · 0:47
You can run the model through the API for just a little over a dollar per million output tokens. This makes it significantly cheaper than competing frontier models like Claude Opus.
03 Can I run the new DeepSeek model locally? 2 sources · 1:45
You can download and own the weights for free forever, but running it yourself requires substantial hardware. Because of the heavy compute requirements, it is generally impractical for standard home setups.
04 How do open weights models compete with proprietary frontier models? 2 sources · 2:16
The release of expensive frontier models triggers a race to build alternatives that match their capabilities at a fraction of the cost. Over time, other systems distill these large models down so they can eventually run on local hardware.
05 Did the model architecture change in this new version? 2 sources · 2:38
The underlying architecture did not change at all for this release. Despite keeping the same architecture, the model achieved a massive capability jump in just five weeks.
06 How did they achieve performance gains without changing the architecture? 2 sources · 3:05
The capability jump comes from learning via self-feedback rather than relying solely on massive new training data. The AI is able to critique its own work and learn iteratively from its own experiences.
07 How does learning from self-feedback work for coding tasks? 2 sources · 3:35
The AI writes code for a game, plays it, and notices bugs like broken collision detection or missing acceleration. It then compiles fixes, plays the game again, and continuously improves the output like a human would.
08 What is the main challenge when an AI judges its own work? 2 sources · 4:28
It is very difficult to implement because the AI essentially judges and rewards itself. Determining whether it is actually performing well during this autonomous process is not always straightforward.
09 How fast are open weights alternatives catching up to proprietary systems? 2 sources · 0:19
Frontier AI models are improving at an incredible pace, and open weights alternatives are right behind them. Major capability leaps can happen in as little as five weeks.
10 What infrastructure is recommended for running and fine-tuning models? 2 sources · 5:05
Cloud GPU platforms provide the powerful hardware needed to reproduce research papers, run inference, or fine-tune models. They allow developers to test ideas and see results in minutes.
Section-by-section summary
4 parts
A surprisingly massive update in a short timeframe
Just five weeks after the previous release, DeepSeek launched a new iteration that delivers dramatically better software building and terminal operation capabilities.
Evaluating true costs using task-based metrics
While advertised as dramatically cheaper per million tokens, token efficiency differences mean cost-per-task provides a more realistic comparison against other frontier models.
Open weights democratize frontier-level capabilities
Releasing model weights enables the broader ecosystem to distill and optimize the technology so it can eventually run locally on consumer hardware.
Self-feedback loops drive capability leaps without new architectures
The massive performance increase stems from learning through self-critique, allowing the model to iteratively test and refine its own work like a human developer.
Connects to your library
by shared topics
n next · p previous clip · t highlights/full
< > slower/faster · / search what was said
-
DeepSeek4× · first 0:00
A newly released AI system and model family offering high performance at low task costs.
-
Claude Opus2× · first 0:58
A frontier AI model used as a performance and cost baseline for token comparisons.
-
Open weights1× · first 4:43
Model weights that can be freely downloaded and owned by users.
-
Self-feedback1× · first 3:05
A learning mechanism where an AI critiques its own work and learns from its experiences.
-
Lambda2× · first 5:05
A cloud GPU provider used for running inference, fine-tuning, and reproducing AI research.
-
Nvidia GPUs1× · first 5:25
Powerful hardware provided by cloud platforms to run experiments and train models.
829 words · about 3 min to read
A surprisingly massive update in a short timeframe
Good news everyone, a new DeepSeek AI system is available. Hooray. So, is this a small or a big jump in AI capabilities? Well, doesn't sound big because the previous version was 4.1 Flash and this one is called 4.1 Flash 1017. I know names. Only about 5 weeks have passed since the last version and they didn't even want to increment the version number. So I guess it's not a big jump, right?
Boy, was I wrong when I thought that. Holy mother of papers. Look at that. It
Evaluating true costs using task-based metrics
1 Highlight · Evaluating true model costs via task-based metrics
builds much better working software and is way better at using the terminal to operate the computer. And it's really good at other things, too. You can run it from the API for just a bit more than a dollar per million output tokens. Wow. But can that really be? This is 15 times cheaper than Claude Opus 5.5. Finally, intelligence too cheap to meter. Well, that sounds a bit better than it really is. You see, we always measure dollars per a million tokens, but different AI systems use a different amount of tokens for the same task. And DeepSeek is not among the best in this. In my independent tests, I saw that it can easily take two to three times more tokens to accomplish the same thing as Claude Opus. So, we should always think in cost per task, not per million tokens. Now, the cost per task for
DeepSeek is still quite good, just a bit less good as what is advertised. Now, here is the best part. You can download and own the weights for free and own it
Open weights democratize frontier-level capabilities
forever. You will need H. Let me check my notes. Ah, yes, exactly $1 to get the hardware to run it. So, I can't run that myself at home. However, that's not the point. The point is that after a new AI system appears, it starts a race to create something close to as good, but at a fraction of its size and cost. Deepseek does that to the frontier models, while other systems distill Deep Seek down into smaller and cheaper systems. So yes, first one bazillion dollars will be needed and maybe less than a year later you find yourself being able to run it on a beefy laptop.
But wait, the architecture did not change at all in this new version. So how on earth did they get such a huge jump in capabilities? Dear fellow
Self-feedback loops drive capability leaps without new architectures
2 Highlight · Achieving massive capability jumps through self-feedback
scholars, this is two minute papers with Dr. Koa. So, how now? I looked at the language they used in their announcement and cross-cheed it against other real research to see what they may have meant. No guarantees though. But I think the key term here is learning from self feedback. What's that? That's the opposite of what everybody is talking about, which is needing to obtain more training data. This is the opposite.
This means that the AI is able to critique its own work and learn from it. That is absolutely amazing. Now, hold on to your papers, fellow scholars, and imagine that it could code up a racing game. Then it starts playing it and notices, wait a minute, the collision detection does not seem to work. Things go through each other. Well, that's not great. Let's fix that. Compile the new game and keep playing and improving it.
Now there is no acceleration. Fix that. Then visual detail is quite poor. Fix that too. And so it goes. So if you remember one thing about this episode, let it be this. We keep talking about giving more and more training data to AI systems to learn from. But previous research works have shown that we can create situations where the AI can learn from its own experiences learning like a human. And maybe this is it coming online in a real product we can own and use. That is incredible and very hard to pull off because the AI essentially judges and rewards itself and it's not always easy to tell when it is doing well or not. The Frontier AI models are
improving at an incredible pace and the open weights alternatives are right on their heels and all this improvement in just 5 weeks. I am stunned. Amazing. It is lovely to see that there is intense competition in this space. That is amazing news for all of us fellow scholars. More intelligence for all of us, cheaper.
What a time to be alive. I use Lambda to reproduce AI research papers often in minutes. It's also great to train your own models or fine-tune an existing one. Run inference or text to image or video. Easy peasy. Running a Deepseek chatbot or agent. Super fast, super reliable. Lambda gives you powerful Nvidia GPUs to run your own experiments. I test ideas from the papers I cover and moments later results. Love it. Seriously, try it out now at lambda.ai/papers.