Bittensor Co-Founder Jacob Steeves Unveils Decentralized Afine 1 Model and Subnet Tokenomics Roadmap
Summarized by AI from reporting by JM Crypto, published under our editorial policy.
At Bittensor's Exploit conference, co-founder Jacob Steeves revealed that 23 subnets are now generating real revenue, showcased the performance of the decentralized Afine 1 model, and detailed upcoming cross-subnet 'gamma tokens'.
Speaking at Bittensor's Exploit conference, co-founder and Afine CEO Jacob Steeves highlighted the network's shift toward self-sufficient commercial utility. Steeves disclosed that 23 of Bittensor's 128 subnets now bring in paying customers and generate direct hard revenue, moving the protocol from an incentive experiment to an operational compute ecosystem. Infrastructure providers within the network have seen significant usage growth, including GPU subnet Liam 7 generating over $2,000 per day in compute rentals alongside active deployment across storage and inference subnets such as Hypius, Shoots, Targon, and NG.
Steeves showcased the release of Afine 1, a 36-billion total parameter (3-billion active parameter) model trained entirely on Bittensor's decentralized storage, compute, and inference stack. Developed competitively by an unorganized, adversarial network of global participants, Afine 1 matched the combined benchmark intelligence score of Anthropic's 4.1 Opus while achieving inference speeds of 500 tokens per second on an H100 GPU. Steeves also highlighted other open-source milestones built on the protocol, including Tutonic 2, a 110-billion parameter decentralized foundation model, and Albdo, an LLM-as-a-judge subnet targeting software engineering benchmarks.
Looking toward 2026 and 2027, Steeves outlined upcoming protocol expansions aimed at driving capital into the network and enabling inter-subnet commerce. Bittensor plans to deploy canonical bridges to track external capital flows and usage credit purchases, incorporating outside demand directly into emission calculations. Internally, the network will introduce "gamma tokens"—fungible usage credits representing infrastructure services—allowing subnets to allocate portions of their token emissions to purchase compute, storage, or inference directly from peer subnets. Steeves concluded by emphasizing that building mechanism-driven, permissionless AI infrastructure remains critical for keeping artificial intelligence open and independent from centralized state or corporate control.