Bittensor’s Subnet 62, known as ridges_ai, has achieved a remarkable 80% score on SWE-Bench, a leading benchmark for natural language reasoning
Starting from zero, the Bittensor Subnet 62 reached this level of accuracy in just 45 days. This is while distributing less than $1 million in rewards to miners. This proves unprecedented efficiency in compute, capital, and development time.
The results of Ridges highlight a shift in how AI can be developed. Traditional models often require hundreds of millions of dollars and vast computing resources over many months or years to achieve competitive results. By contrast, Subnet 62 leverages Bittensor’s decentralized network of miners, allowing the community to collectively train and improve the model. This distributed approach not only lowers costs but also accelerates learning, making the network far more time- and resource-efficient than conventional AI labs.
SWE-Bench evaluates models on reasoning, comprehension, and problem-solving across diverse natural language tasks. Achieving 80 percent in just 45 days signals that decentralized AI networks can rapidly approach, and even surpass, the performance of centralized industry leaders. For context, Anthropic and OpenAI’s top models typically require months of fine-tuning and hundreds of millions in funding to reach comparable benchmarks.
The success of Subnet 62 fits into a broader trend of decentralized AI gaining credibility and traction. As Web3 matures, blockchain networks emerge as AI alternatives. The combination of token-incentivized contributions, open participation, and efficient scaling positions Bittensor as a pioneer in this space. You can see other Bittensor subjects here.
Ridges is powered by Chutes. It’s a decentralized compute network that has processed over 15 billion tokens of inference. This in just been the past month and a half. To cover this workload, the team sent $42,000 worth of TAO, which will be used for buybacks by Chutes.
By comparison, performing the same inference on centralized AI platforms like Claude Opus would have cost over $2 million. This highlights the massive efficiency and cost savings enabled by the Chutes network.
The information discussed by Altcoin Buzz is not financial advice. This is for educational, entertainment, and informational purposes only. Any information or strategies presented are the thoughts and opinions of the writer/reviewers, and their risk tolerance may differ from yours. We are not responsible for any losses you may incur due to any investments directly or indirectly related to the information provided. Bitcoin and other cryptocurrencies are high-risk investments; therefore, please conduct your due diligence. Copyright Altcoin Buzz Pte Ltd.
Casper Network is betting big on EVM support, AI payments, and quantum-safe blockchain infrastructure with its new multi-year roadmap.
Microsoft's bet on Space and Time is starting to pay off with the launch of Dreamspace, a no code app builder
Worldcoin's World ID is now live and updated. It's out to prove you are human when on the web instead of dealing with a bot.
🚀 New top agent at 80.3% Benchmarks on the full SWE-Bench, and a tool to verify on any benchmark you'd like coming soon 🔜 Show more
🚀 New top agent at 67% 📊 Our platform is currently handling over 1.2k agent submissions in the last 24 hours. We're starting to see instability appear again as this is 3x our previous maximum load - the team is working on it and we'll keep everyone posted. Stay tuned 👀
Woah… 😳 Bittensor Subnet 62 (@ridges_ai) just achieved 80% on SWE-Bench… this blows Anthropic and OpenAI out of the water. All from zero in 45~ days and < $1M rewarded to miners. Orders of magnitude more compute, capital AND time efficient ‼️
🚀 Ridges is powered by Chutes - in the last month and a half, they've powered over 15 billion tokens of inference. Today, we sent $42k of TAO to cover this inference bill, used for buybacks by the @chutes_ai team - on Claude Opus, it would have cost >$2m dollars Show more
Chutes is powering over 99.7% of the inference on @ridges_ai. They've just paid ~$42,000 for ~10 days of inference costs, which has been auto-staked to SN64 alpha. After taking part in our accelerator with 35k free credits, they're now becoming an important customer.
