Hey everyone, myself Gatin. I'm DevRel at 0G. Here I'm just giving you a brief introduction on what 0G is and how you can leverage our bounty track of $15,000 during this hackathon. So, first of all, welcome everyone. I'm glad to be here in front of you and good to see like a full full room of audience today.
So, that being said, we'll quickly get started. So, yeah, before I start like how many of you have already heard 0G and already have built with 0G? Oh, that's a good number. Uh yeah, for those who are new to it, I'll just give you a quick introduction and if you have any questions, you can reach out to me later at the booth. We have our colleague as well.
So, yeah, feel free to reach out to us after this session as well. So, uh what 0G is. So, 0G is a layer one chain uh for on-chain AI. Just to give it very brief, we are a L1 EVM compatible blockchain for on-chain AI. That is, we have a complete stack with which you can bring AI on chain.
So, that uh with that uh we have like four different pillars. So, one is the storage, we have the DA layer, we have the compute, and we have a service marketplace. So, with which you can achieve from like model training to having like uh data sets stored on a 0G storage, having them trained uh and also getting like verifiable inferencing for those AI models, and you could use that inferencing for any other dApps that you're building on top of us. So, we are a completely modular stack and I would love to go in deep for like each architecture that we have right now. So, let me quickly start with the chain.
So, we are a EVM compatible blockchain as I mentioned earlier. So, any smart contracts that you have deployed with other chains, you could easily migrate and deploy to 0G by just providing our like chain details, our the RPC endpoints. We are currently on our mainnet. We did our mainnet launch last September. So, if you need any testnet token support, you can just reach out to us at the booth or you could visit faucet.
0g.ai to get testnet tokens. Uh that's on the chain. Uh on top of that, we have like three core pillars. So, first is the storage.
So, we have an inbuilt storage. You might ask, what does like a AI chain doing with having a storage on top of them? And why don't they use other storage providers that are already in the web3 market? So, we require a native storage into us because having AI models and getting verifiable inferencing from it takes up a better latency and also storing like large-scale models on like a decentralized uh distributed storage takes up uh it's like real it's really expensive. So, to just give you a quick example, uh our team recently uh did a research called as Dialect X with which we have uploaded a 100 billion plus parameter model on 0G storage at office data bandwidth.
So, these are like a few uh details that can help you understand like why you need to have like a inbuilt storage. And uh on top of that, you could upload we have you can it supports both structured and unstructured data as data sets as well. So, you could upload data sets, models, assets, vectors, any file. You could just think of it as a decentralized drive storage as well. You could upload any data into it.
All you need to do is like get the root hash and with the root hash, you could also retrieve the files back from the storage. The next is the data availability layer. So, the data availability layer acts like an off-chain storage or like a memory layer where you have like a provenance record of all the previous history what happened in the blockchain. So, uh I'll just dive deep about DA and storage in upcoming slide, but this is just a quick brief. Moving on to compute, so this is like the cream on top of the cookie for 0G.
So, with that, you will be able to do like verifiable inferencing from AI models. So, for to give you an example, we have models like GLM-5, we have like DeepSeek, GPT-3, Gwen. So, these are a few models on top of 0G and you will be able to get verifiable inferencing from it. So, you might ask like what do you mean by verifiable inferencing? It means that the output the input that you give to the model and the processing of the output by the model happens in a TE environment.
That is like a trusted execution environment. So, there is no like external interference when like the model processes the output. For example, uh you might still have a question like why it's required. For example, you're using a centralized provider, it's like complete black box. You aren't sure like what is happening behind it, what is the censorship resistance mechanism that is there behind it as well.
Uh so, for example, if you are asking some product recommendation and if suggests that this is the product X and this is like the best in the market, so you might not know whether that company X paid them a ton of money to promote it out over like the model's actual trained data. So, to avoid all that, there is a big transparency aspect required and uh for that, that is where we solve it by having like a TE environment and having the data processed in that. On top of that, uh yeah, in the inferencing and we also have like fine-tuning live on mainnet. So, with these existing models, you can fine-tune them on chain based on your requirements. So, that is live as well.
And moving on top of that, we have the service marketplace. So, this isn't uh this is just briefly live yet. In this, you will be able to register yourself as a provider. That is, you can like register your GPUs, you can contribute to the network, and anyone who wants to deploy their models can deploy on top of that and you can get paid based on the contribution that you give. So, just like a marketplace on top which leverages the complete stack.
So, as I mentioned like these are the four core components. Just to give you a brief example of as I mentioned that the storage is much cheaper. Just to give you an example, AWS S3 it costs you like $100,000 per month for like the storage, but whereas 0G storage it's going to be like $5,000 per month. This is like a quick comparison on what it takes to store at S3 compared to 0G storage which we have approximated up to like 95% cost savings. So, regarding the inference and training, we as mentioned, we have the the largest model that we have is GLM-5 which is 744 billion parameter model.
And uh as mentioned, we have like the trusted execution environment for the processing. And also like the output that you get from the model like the TE the model is in the TE and the processing happens. And when you get the output, just to give you a surety that it is verified, the you also get like a ZKML proof that it was actually processed in a TE environment. So, this is regarding the DA and chain. So, the DA throughput is 50 Gbps which is faster than all the other DA that is out there in the market.
And our chain has achieved like 11K TPS 11K transactions per second per shard. And also the block finality is 1 second which we have also achieved right now. We haven't updated it it yet. It's almost like 500 milliseconds per block time as well. So, it's really instantaneous for like the transactions that you can make.
So, that it's like almost you interacting with a centralized provider. There will not be any like big latency for you to get output from these models as well. The one other core feature that we have is we have a standard called we our team invented standard called as ERC-7857 which means it's called as INFTs, intelligent NFTs, or agentic IDs. What it does is for all your AI agents that you build, this gives like an identity to it. So, it's like an NFT bound to that agent.
So, it what it means like if you transfer this NFT to some other wallet address, that means that you have transferred the ownership of the agent to that address. So, this becomes this becomes like a core feature. So, for example, in a previous hackathon in the zero coding event, so uh which we held like the day before, so that was a team that built like we have the team members here as well. So, he built like an X402 payment standard between these agents and that standard like the payment takes place between those agentic IDs. So, you can just think of it like an ENS or like an identity to that agent.
So, you could try building on that as well. And one other thing that we would like you to like you to build is using like the X402 standard. Uh this is a quick brief on our mainnet stats. We have like 300 plus uh ecosystem partners. Proud to say that we are one of the largest on-chain AI uh ecosystem.
And also we have like 400 plus integrations across the stack. And we had like well-funded. We have around like $325 million funding. So, what it can relate to you is that we have a subgrant amount of like $88.8 million.
And in that, we have an ecosystem support program which accounts to around $8.8 million. It is called as guild.0gfoundation.ai.
It's guild on 0G program. What you can do is if you have any idea and you want to build on top of 0G and you need consistent support whether it may be technical, monetary support, or whether it may be with any of our ecosystem project connections that you require, we provide that in this program. So, I'll share the link post this session and please do apply if you are like maybe if you build with 0G here and you want to scale it from here, that might be a good place for you to get like a proper support channel. This is Alignment notes. I'll just give you a brief we just did a note sale earlier in 2025 December.
So, with which we have given like Alignment notes to our community members. So, they will have control or like they have the voting rights to what happens on 0G chain in the future days. Like that's like a very quick brief. But that comes into a lot more in-depth details where when it's like completely on chain and you have like a whole supply chain running on chain. So, this gives like a proper monitoring for like the community members to like reject if there is any malicious activity on chain.
So, yeah, these are like the problems that with AI today which I mentioned earlier that is like the lack of trust. Vendor lock-in like you'll be locked into vendor for example like AWS S3 for example. You're just really locked into that vendor. And there is no ownership. That's like a centralized control.
That's just like a quick brief. Yeah. I'll just give you a quick rundown on what we have at the ETH Global. So, that's like a quick rundown of what 0G is. We just have a quick time crunch so I'm just trying to wrap it up as fast as I can.
So, we have like a $15,000 in total for the bounty. And we have $6,000 for the best open claw agent that can be built on top of 0G. $6,000 for a DeFi app and $3,000 is like a wild card track. So, you can build anything on top of 0G which is not related to the first two tracks. On the ETH Global side you can find in detail for like also the examples that we would like you like to see you build during this hackathon.
That is for the open claw. I'll just give you a quick brief and this is for like the DeFi app on 0G. And this is for the wild card track. We have a $3,000 allocated for this as well. If you have any other ideas other than the first two tracks, you could apply for this track.
Yes. So, the first QR code helps you to get like if you are a really good wipe if you're into wipe coding. Hopefully most of you might be into it right by now. So, that has like our complete documentation all the resources that we have. We have made it to wipe we have made it wipe coding friendly.
So, we have like compute we have skills for like all this complete stack. We have cloud skills. We have like We have cloud skills. We have like starter kits for all our toolings. Our complete docs has like an AI context page which you can feed to your LLMs and it'll be able to get like the complete context of like the 0G's stack and you'll be able to build just by wipe coding.
And the next one is like the builder hub. It's build.0g.ai. It has all the resources in one place.
You can get all the zero coding link in it as well. The one important thing that I would like you to do is join this TG group if you're planning to build on 0G during this ETH Global. So, and we also have a channel for faucet request. So, if you need any faucet token just drop your address in that and we will be we we will be able to help you with it. And also if you have any questions or you want to meet other people who are building on top of 0G as well, that is a good channel for you to be in.
So, please do join the Telegram group and we would be able to help you with that. And that's mostly it. If you have any questions, so feel free to like we have like a minute or so. So, yeah, please. Does that like still qualify for the second bounty for example?
Like if we are using the chain but then the final result will be on EVM when there's like more lending protocols and and swaps and so on. Yes, I understand. So, we need to use multiple bounties as well. So, 0G stack is completely modular and you can utilize like we don't have like a requirement hard requirement that you need to use everything on top of 0G. The more you use the better it is based on your use case.
But for your use case if you think that just the inferencing should be fine and you can best leverage like maybe the thing that you mentioned with other lending protocols and other chains, then feel free to do that. So, we don't have like you need to completely use 0G stack. It's modular and you can use the aspects based on your requirements. Okay, thank you.
Yeah. In your slides you mentioned rag. How does that work? Where is the vectorized database?
Currently we don't support rag. We are it's like completely internal right now but we will be soon launching rag on chain as well. Cool. So, we could do a different vectorized database and then use your inferences. Correct.
Currently we only have like the model inferencing and also if you want to host your own model you need a H100 or H200 GPUs. But other than that with the existing models with the fine-tuning part is live right now. Thank you. Yeah. Awesome.
Thank you so much everyone. If you have any questions, please do drop by at our booth. Thank you.
Automatic transcript — names and jargon may be misspelled.