MineBench - efficient computing power

What if benchmarking your CPU actually rewarded you? Run a benchmark. Get your score. Earn solana:67ipDsgK6D7bqTW89H8T1KTxUvVuaFy92GX7Q2XFVdev . This is MineBench.

Networking, but with an AI superconnector. Today, @boardyai helped MineBench explore potential connections and collaboration opportunities with @hf0. Building a startup is not just about writing code. It's also about finding the right people, opening the right doors, and being in the right conversations. Great to see AI making that process smarter and more accessible. The future of networking is already here.

Why does MineBench use your PC, Monero, and mining under the hood? Most people own a powerful computer but use only a fraction of its potential. Meanwhile, running AI, verifying hardware performance, and accessing computing resources can be expensive. We're building MineBench to connect these worlds. Your PC is more than a device for gaming, work, or browsing. It is a source of computing power. MineBench makes that power useful through CPU benchmarking, stress testing, and a mining-based reward system. Our goal is to give users a practical reason to participate, contribute computing resources, and earn rewards. Why PCs? Because accessible computing matters. Not everyone can afford an expensive GPU workstation or dedicated mining hardware, but millions of people already own capable CPUs. By building around consumer hardware, MineBench lowers the barrier to participation and helps create a more distributed computing ecosystem. Why Monero ($XMR)? Monero is a privacy-focused cryptocurrency secured by Proof of Work. Its RandomX algorithm is designed for general-purpose CPUs, making ordinary computers meaningful participants in network security. This aligns with our philosophy: computing participation should not be limited to large industrial mining farms. Learn more: getmonero.org/get-started/mi… What does Monero mining actually do? Mining is not simply generating coins. Miners compete to produce valid Proof-of-Work solutions, securing network consensus, helping prevent double-spending, and allowing transactions to be confirmed without relying on a central authority. Miners receive block rewards and transaction fees for contributing to network security. How does mining relate to private transactions? Monero uses privacy technologies such as ring signatures, stealth addresses, and RingCT to protect transaction details. Mining secures the blockchain on which these transactions are recorded, but it does not itself make a transaction private. Privacy and mining solve different problems, yet both are essential parts of Monero's design. What about RandomX v2? The evolution of RandomX is important for the future of CPU-oriented mining. RandomX is designed to make general-purpose hardware competitive and resist the dominance of specialized ASICs. For MineBench, this direction is particularly relevant because we believe consumer CPUs should continue to have a meaningful role in decentralized computing. Network compatibility and miner software must follow the actual consensus rules of any upgraded network. Why put mining under the hood? Because we want rewards to be connected to real participation rather than meaningless clicks. Mining provides an established mechanism for contributing computing power to a decentralized network. MineBench integrates this mechanism into a broader user experience that includes benchmarks, stress tests, and performance tracking. The idea is simple: make participation useful, measurable, and rewarding. What does the user get? Users gain access to CPU performance benchmarks, hardware insights, stress testing, and an opportunity to earn rewards through eligible participation. They can better understand their hardware while contributing to a distributed, community-driven ecosystem using computers they already own. Mining consumes electricity and generates heat, so users should remain in control of resource usage and understand the associated costs. What is solana:67ipDsgK6D7bqTW89H8T1KTxUvVuaFy92GX7Q2XFVdev for? BMT is the utility and reward token of the MineBench ecosystem on Solana. Our economic model connects mining revenue with the token economy: mined XMR can be converted into SOL, used to acquire BMT, and allocated toward user rewards and treasury sustainability. This creates a link between measurable computing participation and ecosystem incentives rather than relying solely on token emissions. Rewards and token value are not guaranteed. The bigger picture: MineBench starts with something people already have - a PC. Monero provides a CPU-oriented Proof-of-Work foundation, RandomX supports our focus on accessible computing, and BMT connects participation with the MineBench reward economy. We want to build an ecosystem where hardware contributes, users benefit, and the community helps drive sustainable growth. We are building for the long term. minebench.xyz @monero @getmonero #Monero #RandomX #MineBench

Why does the world need a decentralized AI computing network? AI is growing rapidly, but access to computing power remains one of the biggest barriers to adoption. Running powerful AI models locally requires expensive GPUs, large amounts of RAM, high-performance hardware, and significant electricity costs. For many individuals, developers, and small businesses, investing thousands of dollars in equipment just to access AI capabilities is simply not practical. At the same time, thousands of GPUs and CPUs around the world remain underutilized. People invest in expensive hardware, but it often sits idle when they aren't using it. This is an enormous amount of untapped computing potential. We believe there is a better way. $MBIQ is building a decentralized AI computing network that connects hardware owners with people who need computing power. Instead of everyone having to purchase and maintain their own expensive infrastructure, users can access computing resources from a global network of independent providers. For users, this means more affordable access to powerful open-source AI models, without the need to invest in expensive hardware or maintain local infrastructure. It creates more choice, competition, and freedom from dependence on a handful of centralized providers. For providers, it creates an opportunity to monetize idle GPUs and CPUs. Hardware that would otherwise sit unused can generate income by processing real AI workloads. Every provider becomes part of a global computing infrastructure. What could this change for the AI market? We see the potential for a fundamental shift in how AI computing resources are distributed and monetized. Instead of infrastructure being controlled exclusively by large corporations, independent hardware owners could become an important part of the global AI economy. More providers mean more competition, greater resource availability, and potentially lower computing costs. We don't believe centralized data centers will disappear. We believe they should have an alternative. Our vision is to make AI computing more accessible, turn unused hardware into productive resources, and create an ecosystem where both users and providers benefit. Don't just own computing power. Put it to work. This is what we are building with $MBIQ. #DePIN #AI #Solana

What is AI Quantization, and why does it matter? Quantization is a technique that reduces the precision of AI model weights by representing them with fewer bits. This makes models smaller, reduces memory requirements and can improve inference speed, allowing larger models to run on more affordable hardware. But there's a trade-off: model quality. How does it affect AI performance? Q8: Usually retains almost all original model quality. Q6–Q5: Generally very good quality with minor degradation. Q4: A popular balance between model size, performance and quality. Q3–Q2: More aggressive compression, with a higher risk of losing accuracy and reasoning capabilities. Q1: Extremely aggressive compression that can significantly affect reliability. Let's look at a simple example. Imagine showing the same image to an AI model and asking: "Describe what you see." Depending on quantization, possible responses might look like this: Q8: "A golden retriever is sleeping on a rug beside a red ball." Q6: "A golden retriever is resting next to a red ball." Q4: "A dog is sleeping near a toy." Q3: "A brown dog is lying on the floor." Q2: "A dog is playing with a ball." Q1: "A cat is sleeping on a sofa." These are illustrative examples, not actual benchmark results. Quantization doesn't guarantee these errors, but aggressive compression can increase the risk of losing details, accuracy and reliability. Why should you care? The smaller the model, the less memory it requires. But smaller doesn't always mean better. The challenge is finding the right balance between model size, inference speed (tokens per second) and output quality. At MineBench IQ, we believe understanding these trade-offs is essential to making powerful AI more accessible and using computing resources more efficiently. MineBench IQ - making powerful AI accessible to everyone.

Powerful AI shouldn't require everyone to own expensive hardware. Everyone wants to run advanced AI models locally. But there's a reality that marketing often overlooks. Running an AI model isn't just about downloading it. You need to understand model size, quantization (Q4, Q5, Q8 or FP16), memory requirements, and actual inference performance measured in tokens per second. A model running locally doesn't necessarily mean a fast or usable experience. There's another problem we rarely talk about. Companies and individuals invest thousands of dollars in professional AI hardware, such as NVIDIA RTX PRO 6000 GPUs. Yet these expensive resources can sit idle for hours or even days when they're not being used. Meanwhile, someone else is saving money to purchase exactly the same hardware, only to have it sit idle when their own workloads are finished. We're duplicating expensive infrastructure instead of using what already exists. This is what we're building with MineBench IQ. A distributed AI compute marketplace that connects available computing resources with people who need them. Have a data center or powerful multi-GPU infrastructure? Become a provider. Host large models such as DeepSeek, Kimi or Qwen, share your inference capacity, offer different performance tiers measured in tokens per second, and earn from resources that would otherwise remain unused. Have a powerful gaming GPU or workstation? You can contribute too! Run smaller language models, coding assistants, image-generation models such as FLUX, and other AI workloads. You don't need enterprise infrastructure to become part of the network. What if you need access to powerful AI? You don't need to buy expensive hardware yourself. Access models and computing resources from providers across the network. Need Qwen for coding? Access a provider running it. Need DeepSeek for reasoning? Use infrastructure capable of hosting it. Want to generate images with FLUX? Connect to a provider with suitable resources. Need more memory or higher inference throughput than your local machine can provide? Access larger computing infrastructure. Our vision is to make better use of the hardware we already have. Providers monetize idle computing capacity. Users get access to powerful AI without purchasing expensive equipment. Developers get access to a distributed pool of computing resources. Instead of everyone buying the same expensive hardware and leaving it unused, we can build a shared ecosystem where computing power is available when and where it's needed. MineBench IQ - turning idle computing power into efficient computing power.

We're experiencing further issues with GitHub, so we've decided to move our entire codebase to Bitbucket.

A benchmark token is not a bubble map token.

Open Source model

Powerful AI hardware often sits idle or runs below its full potential, while demand for AI computing continues to grow. Join us as a provider, and let’s build our new app together. Join our Discord to get involved.

Looking at the price of the new Mac Studio M5 Ultra 512GB and its RAM capacity, we realize that we need at least 100 of these in our own decentralized network. Who’s ready to join? Our Discord is open to miners and AI providers.

What defines a good token?