Bittensor Weekly Recap – 2 February 2026: News, TAO.
Big Events in the Bittensor Ecosystem
Mentat Minds has officially released its new Mentat SoS (Sum of Subnets) index, offering Bittensor investors a streamlined way to gain exposure to the entire subnet ecosystem with just one click.
What is the Mentat SoS Index and Why It Matters
The Bittensor network continues to attract innovative financial products as its decentralized AI infrastructure grows. The newly launched Mentat SoS index aggregates performance across multiple Bittensor subnets, eliminating the need for investors to manually select and manage individual subnet positions.
The Sum of Subnets index functions as a diversified investment vehicle within the Bittensor ecosystem. Rather than betting on a single subnet’s success, users can now spread their exposure across the network’s expanding range of AI-focused subnets through a single transaction. This development addresses a common challenge faced by Bittensor participants: the complexity of evaluating and investing in dozens of specialized subnets, each with unique use cases ranging from text generation to data scraping and machine translation.
As Bittensor matures, products like the Mentat SoS index signal growing institutional-grade infrastructure around the TAO token ecosystem. Simplified investment tools lower the barrier to entry and may attract broader participation in the decentralized AI network.
The Mentat SoS index is now available through Mentat Minds’ platform.
Subnet Updates
Synthdata (SN50) Releases Synth API and Subscription Plans for Probabilistic Market Forecasting
Synthdata has released the Synth API together with a set of subscription plans, expanding access to its probabilistic price forecasting product. The API is intended for use in prediction markets, options trading workflows, and decentralized finance applications, areas that increasingly intersect with intelligence-oriented systems such as Bittensor.
The Synth API differs from conventional market data tools by providing probability distributions instead of single price targets. These outputs are designed to be consumed programmatically, allowing developers and quantitative traders to incorporate distribution-based forecasts into automated decision logic. This approach reflects a broader trend in crypto and DeFi toward data structures that can be integrated into autonomous systems, including those influenced by Bittensor-style architectures.
According to the company, the API has been tested in live environments prior to release. These tests included internal prediction market strategies, options-related workflows, and DeFi use cases focused on risk evaluation. Reported applications involved assessing implied market confidence, comparing expected volatility with market pricing, and adjusting execution rules based on probability-weighted scenarios.
In prediction markets such as Polymarket and Kalshi, the Synth API can be used to derive alternative probability estimates for contract outcomes. By comparing market-implied probabilities with distribution-based forecasts, users can evaluate discrepancies in range width and dispersion. This type of analysis aligns with data-driven decision models often discussed in the context of Bittensor, where forecasting quality is treated as a measurable input rather than a subjective view.
Options-related use cases include analyzing expected realized volatility, downside probability, and the likelihood of price range breaches. These metrics can be applied to trade filtering, sizing logic, and risk constraints for strategies involving short-dated or multi-leg options. The API is positioned as an input layer rather than a strategy engine, leaving execution logic to the user.
Within DeFi, Synthdata frames the API as a source of forward-looking risk parameters. Potential applications include adjusting leverage thresholds, defining probabilistic liquidity ranges, and informing automated hedging logic for perpetual positions. Such use cases are consistent with ongoing experimentation around autonomous and adaptive systems, including those associated with Bittensor-influenced research and development.
The company has introduced three access levels. The Free tier provides dashboard access without API calls. The Standard tier, priced at $49 per month, includes a one-time allocation of 50 API calls and is intended for testing and early development. The Pro tier costs $199 per month and includes 500 API calls per month, with the option to purchase additional usage.
Synthdata reports that early users include developers building automated prediction market tooling, options monitoring dashboards, and DeFi logic that updates parameters based on forecasted volatility. These implementations reflect a broader shift toward probabilistic inputs in automated systems, including those developed in parallel with Bittensor-related ecosystems.
The Synth API is now available to the public, with documentation and access details provided through Synthdata’s official channels.
Crunch Launches Mining Operations on Synth (SN50) Subnet Within the Bittensor Network
Crunch has announced the start of mining operations on the Synth price prediction subnet on Bittensor, marking its first mainnet deployment of a Coordinator in a decentralized intelligence network. The initiative brings Crunch’s quantitative research community into the Synth ecosystem, where contributors generate probabilistic financial forecasts evaluated on statistical accuracy.
According to the announcement, Crunch will deploy forecasting models developed by its global community of more than 11,000 machine learning engineers and over 1,200 PhD-level researchers. These models will produce probabilistic price distributions for cryptocurrencies, tokenized commodities, and equities, contributing directly to the Synth subnet’s forecasting output.
Synth operates as a decentralized prediction engine on Bittensor, rewarding contributors for producing accurate probability-based forecasts rather than single price targets. The network evaluates submissions across short-term horizons, including one-hour and twenty-four-hour intervals, using standardized scoring methods designed to measure forecast quality.
Crunch stated that its participation focuses on applying collective intelligence to financial forecasting challenges such as volatility clustering, tail risk, and correlation breakdowns during market stress. Individual model outputs are aggregated into ensemble forecasts, which are then used within the subnet’s broader prediction framework.
The collaboration also introduces a forecasting competition covering assets including Bitcoin, Ethereum, Solana, tokenized gold, and selected tokenized equities. Forecasts generated through this process are incorporated into the Synth network, with rewards distributed based on performance.
The mining operation is now live, expanding the contributor base of the Synth subnet and adding institutional and academic modeling expertise to the Bittensor-based forecasting system.
Ridges AI (SN62) Joins Forces With Latent Holdings to Accelerate Development on Bittensor
Ridges AI has announced that it is joining forces with Latent Holdings, a move aimed at accelerating product development and market delivery within the Bittensor ecosystem. The decision follows a year of work on incentive mechanisms, open-source agents, and secure validation infrastructure across the Ridges subnet.
According to the announcement, Ridges identified limitations in attempting to scale product development and subnet operations simultaneously. While progress continued on the technical side, the team stated that competing effectively in the fast-moving AI software engineering market required additional resources and operational experience specific to Bittensor-based systems.
Rather than continuing organic scaling, Ridges opted to collaborate with a larger team that has prior experience operating within the network and bringing products to market. The company described the decision as a strategic response to bandwidth constraints created by ongoing subnet maintenance, hiring, and product shipping.
Under the new arrangement, the subnet will continue operating without changes to its incentive mechanism or alpha structure. Both teams will work together as a single, expanded group, with the stated goal of increasing development speed and delivery quality. No immediate protocol-level changes were announced.
Ridges characterized the collaboration as a step toward improving execution rather than altering the subnet’s core design. The company emphasized that existing mining operations would continue as normal while development efforts shift toward faster iteration and market deployment.
The announcement positions the partnership as part of a broader effort to remain competitive within the Bittensor ecosystem, where teams are increasingly focused on translating subnet research into usable products.
Templar (SN3) Launches “Crusades” Competition to Optimize Training Code on Bittensor
Templar has announced the launch of Crusades, an open competition focused on optimizing large-scale model training code within the Bittensor ecosystem. The initiative follows the completion of Covenant72B, a 72-billion-parameter model trained across a decentralized global network, which has now entered post-training.
According to the announcement, Templar is shifting its focus from model training to improving the underlying primitives that determine training performance. Future runs are expected to place greater demands on the network, including faster gradient aggregation, improved memory efficiency, and more effective use of heterogeneous hardware.
Crusades is designed as a winner-take-all competition for training code optimization. Participants submit implementations such as kernels or training loops, which are evaluated by validators under identical conditions. Submissions are ranked based on tokens processed per second, with the fastest implementation receiving the reward.
Templar stated that optimization techniques discovered through the competition will be integrated directly into its infrastructure. This structure is intended to create continuous pressure toward more efficient training code, with improvements shared across the network rather than remaining isolated to individual contributors.
To maintain fairness, submissions remain private until evaluation is complete, and entries deemed too similar to existing code are rejected. A small submission fee is used to limit spam and encourage original work. Emissions associated with the project now flow through Crusades, and the leaderboard is currently empty as the competition opens.
The project highlighted Crusades as a potential benchmark challenge for large contributor groups operating on Bittensor, including communities recently entering mining activity. Templar described the competition as a test of whether collective intelligence can consistently produce higher-performance training code for frontier-scale models.
The Crusades codebase is publicly available, and the competition is now open to participants.
Kimi K2.5 Becomes Available on Chutes (SN64) as Agentic Models Expand on Bittensor
Chutes has made Moonshot AI’s Kimi K2.5 available on its platform, adding a multimodal, agent-oriented model to its Bittensor-based inference stack. The model supports visual inputs and parallel agent execution, including a swarm mode with up to 100 sub-agents.
According to Chutes, Kimi K2.5 has been evaluated on agentic and software engineering benchmarks such as Humanity’s Last Exam, BrowseComp, and SWE-bench Verified. The deployment allows users to run the model through Chutes’ hosted interface without self-hosting infrastructure.
The release reflects ongoing efforts to make large open models accessible through decentralized compute layers built on Bittensor.
IOTA (SN9) Reports 4.3 Training Loss in Distributed “Train at Home” Model
IOTA (SN9) announced that its Train at Home model has reached a training loss of 4.3, indicating continued convergence in its distributed learning experiment. Loss measures how closely a model’s predictions match expected outputs and is expected to decrease as training progresses.
The model is trained on consumer-grade hardware operated by participants with no machine learning background, using a distributed coordination framework. According to IOTA SN9, this result demonstrates that decentralized training can achieve measurable learning progress outside traditional centralized setups.
The project confirmed that additional participant invites are being rolled out as training continues.
New Subnet Launches:
blockmachine (SN19)
Blockmachine is a Bittensor subnet providing decentralized RPC and archive node infrastructure. Its purpose is to deliver blockchain RPC access with enforced correctness, measurable performance, and predictable pricing.
The subnet separates service pricing from token volatility by pricing usage in USD-denominated compute units, while settling incentives on-chain in TAO. This design supports production RPC workloads and is intended to expand to additional blockchain networks over time.
Sparket.AI (SN57)
Sparket.AI (SN57) is a Bittensor subnet focused on decentralized odds origination and outcome verification for sports and live events. Its role is to produce early, original probability estimates and establish a verifiable consensus signal that replaces centralized sports data and “truth” sources.
The subnet evaluates predictions against realized outcomes and reference market data, rewarding information-leading signals rather than copied consensus. It operates as the decentralized intelligence layer for Sparket’s existing production system, enabling direct integration of subnet outputs into real-world sports data products.
TAO Market Update
Price: $190-$290
Weekly result (7d): -6% to -9%
Ranking: #34-#53
Market Cap: $2.0B-$3.1B
24h Volume: $70M-$175M
As of today, TAO is trading approximately in the $190-$290 range, reflecting current market volatility and active trading participation. Latest market data shows TAO ranking between roughly #34 and #53 by market capitalization, with an estimated valuation of about $2.0B-$3.1B. The 24-hour trading volume remains elevated, roughly in the $70M to $175M range, indicating sustained liquidity and ongoing engagement from market participants.
Top Gainer Subnet: Bounty Hunter (SN20)
Bounty Hunter (SN20) on Bittensor, known as BitAgent, is a decentralized AI subnet where miners compete by solving tasks with tool-enabled language models.
Built on the Bittensor protocol, it rewards high-quality AI performance through on-chain validation and token incentives.
AIBoards acts as the bounty marketplace, allowing developers and creators to complete crypto and AI challenges for rewards. Together, they demonstrate how Bittensor enables scalable, incentive-driven AI development in a decentralized ecosystem.
Weekly change: +36.19%
Price: $0.969935
Market Cap: $6.31M
Volume (24h): $716.87K
