SN81 miner burn down to 78%
The subnet burns 78% of its miner emission now (was 90%): miners are being paid again.
Detected 12h agoView subnet →
Bittensor subnet news
Subnet teams' own channels, X posts, press, newsletters and events read from the chain — GitHub lives in its own tab
SN81 miner burn up to 90%
The subnet now routes 90% of its miner emission to burn (was 69%). Miners earn little or nothing from it while this lasts.
Detected 2d agoView subnet →
Reliquary | Bittensor SN81 τ@reliquary_aiThat rate also shows the network scaling. Emissions are fixed, so more GPUs means more tokens for the same budget. On Qwen3.8-27B, it went from 0.69α to 0.36α per million tokens in two days. Every GPU that joins makes post-training cheaper. Cheaper post-training brings more jobs. More jobs pay more GPUs. Post-training, open to every GPU in the world.
Reliquary | Bittensor SN81 τ@reliquary_aiA centralized lab can only use the GPUs it owns or rents. Reliquary can use any GPU in the world. That’s why we build on Bittensor. Here’s the key. Inference means low latency and no interruptions: an answer can’t stop halfway. Post-training doesn’t need either. Nobody is waiting on a rollout. If a GPU drops mid-task, nothing breaks. So any GPU can contribute. Idle at night, free between experiments. Join for an hour, leave, come back. The work still counts. A centralized lab pays for its GPUs 24/7, busy or not. We only pay for work that’s delivered and accepted. And every GPU owner can know what they’ll earn before they start. Our API now publishes, live, what each task pays per million tokens. Check your card’s throughput, and you have your number. reliqua.ai/api/sft/pay
Linkreliqua.aireliqua.ai/api/sft/payValidator TAO.app (5CoZ…bifj) scheduled a child-key change on 125 subnets
Validator hotkey 5CoZxgtfhcJKX2HmkwnsN18KbaT9aih9eF3b6qVPTgAUbifj announced new child keys on SN1, SN2, SN3, SN4, SN5, SN6, SN7, SN8, SN9, SN10, SN11, SN12 and 113 more, applying at block 9,281,605 (~24h). Stake delegated to it there routes through its children after that.
ValidatorTAO.appSubnetsSN1, SN2, SN3, SN4, SN5, SN6, SN7, SN8, SN9, SN10, SN11, SN12, … (125 subnets)
Between blocks 9,274,365 and 9,274,433
Reliquary SN81 tightens its Bittensor training data exports
Reliquary SN81 filters unfinished answers from training data exports and adds optional deduplication. The Bittensor update puts useful output under scrutiny.
Tao Outsidersource ↗
Validator Tensorplex Labs (5E4z…ULde) re-scheduled a child-key change on 22 subnets
Validator hotkey 5E4z3h9yVhmQyCFWNbY9BPpwhx4xFiPwq3eeqmBgVF6KULde announced new child keys on SN33, SN71, SN72, SN73, SN74, SN75, SN77, SN78, SN79, SN80, SN81, SN84 and 10 more, applying at block 9,265,716 (~24h). Stake delegated to it there routes through its children after that. It replaces the changes that were pending on SN33 (block 9,265,491), SN71 (block 9,265,347), SN72 (block 9,265,350), SN73 (block 9,265,353), SN74 (block 9,265,356), SN75 (block 9,265,359) and 16 more, which will not apply then.
ValidatorTensorplex LabsSubnetsSN33, SN71, SN72, SN73, SN74, SN75, SN77, SN78, SN79, SN80, SN81, SN84, … (22 subnets)
Between blocks 9,258,514 and 9,258,581
Validator Tensorplex Labs (5E4z…ULde) scheduled a child-key change on 22 subnets
Validator hotkey 5E4z3h9yVhmQyCFWNbY9BPpwhx4xFiPwq3eeqmBgVF6KULde announced new child keys on SN71, SN72, SN73, SN74, SN75, SN77, SN78, SN79, SN80, SN81, SN93, SN98 and 10 more, applying at block 9,265,314 (~24h). Stake delegated to it there routes through its children after that. It replaces the changes that were pending on SN80 (block 9,261,738) and SN93 (block 9,263,246), which will not apply then.
ValidatorTensorplex LabsSubnetsSN71, SN72, SN73, SN74, SN75, SN77, SN78, SN79, SN80, SN81, SN93, SN98, … (22 subnets)
Between blocks 9,258,111 and 9,258,178
SN81 miner burn moved from 42% to 69%
Share of the subnet's miner emission routed to burn: 42% → 69%.
Detected 7d agoView subnet →
SN81 miner burn moved from 72% to 42%
Share of the subnet's miner emission routed to burn: 72% → 42%.
Detected 7d agoView subnet →
SN81 miner burn down to 72%
The subnet burns 72% of its miner emission now (was 91%): miners are being paid again.
Detected 7d agoView subnet →
SN81 miner burn up to 91%
The subnet now routes 91% of its miner emission to burn (was 81%). Miners earn little or nothing from it while this lasts.
Detected 8d agoView subnet →
SN81 miner burn moved from 55% to 81%
Share of the subnet's miner emission routed to burn: 55% → 81%.
Detected 8d agoView subnet →
SN81 miner burn moved from 82% to 55%
Share of the subnet's miner emission routed to burn: 82% → 55%.
Detected 8d agoView subnet →
Reliquary | Bittensor SN81 τ@reliquary_aiAffine's inference-verification rewrite takes inspiration from Reliquary. Batch generation is the start. RL on demand, training and evaluation come next All in the same app
Quoting @const_reborn
What the Reliquary team has invented for a minable objective is truly one of the most compelling structures i have seen for a scalable post training mechanism on Bittensor. In the same way that conjectures 66 searches math proofs, Reliquary has miners build compute fleets which search for rollouts on a target model which partition the reward landscape. You spin up a miner, download the network model and scale your inference engine + innovate on your search technique to uncover negative samples from the environments. Those samples become optimal inputs to the training process for the network model. Like how Bitcoin miners bleed energy for a hash. Reliquary miners bleed energy for a rollout. It’s a search problem at heart which is where Bittensor’s abstraction over the search for search algorithms really shines. Reliquary was created by two young engineers with no credentials and no allegiances to anything except a vision for decentralized intelligence. Bravo Rom and Grizz
Reliquary | Bittensor SN81 τ@reliquary_aiMajor subnet teams are already working with us on their training data. Early access stays small so we can work through the first batches with each team. Bring your model, your tasks and the volume you need. If your subnet needs data, join the waitlist:
Reliquary | Bittensor SN81 τ@reliquary_aiA training dataset doesn't need an instant response. Give the network a model, a task and a target volume. Independent GPUs do the work in batches: prompts, multi-turn tasks or full agent sessions. You trade latency for price.
Reliquary | Bittensor SN81 τ@reliquary_aiWe're our own first user. Our network collects coding-agent sessions on SWE-smith tasks in real Python repos. The agent reads files, runs commands, edits code and tests the patch. All of that becomes training data. We'll use those full sessions to train Teutonic.
Reliquary | Bittensor SN81 τ@reliquary_aiThe app is live. You can watch what Subnet 81 is doing, browse task environments and compare GPU prices across providers. No account needed to look around.
Reliquary | Bittensor SN81 τ@reliquary_aiReliquary is building post-training on demand. The network behind our own models is opening up to Bittensor subnet teams. First up: Batch generation. Training data built around your model and tasks. Early access: app.reliqua.ai/early-access
Linkapp.reliqua.aiapp.reliqua.ai/early-accessValidator Astrid (5ELR…qwVh) scheduled a child-key change on 28 subnets
Validator hotkey 5ELREhApbCahM7FyGLM1V9WDsnnjCRmMCJTmtQD51oAEqwVh announced new child keys on SN6, SN8, SN9, SN10, SN20, SN21, SN23, SN30, SN37, SN44, SN47, SN56 and 16 more, applying at block 9,220,938 (~24h). Stake delegated to it there routes through its children after that.
ValidatorAstridSubnetsSN6, SN8, SN9, SN10, SN20, SN21, SN23, SN30, SN37, SN44, SN47, SN56, … (28 subnets)
Between blocks 9,213,717 and 9,213,777
Validator Crucible Labs (5CAQ…7FmJ) scheduled a child-key change on 5 subnets
Validator hotkey 5CAQAgXxoU3NPo1YG5VQKz4Ackh7goNx3TvPD9A6LoqA7FmJ announced new child keys on SN8, SN44, SN76, SN81, SN91, applying at block 9,211,290 (~24h). Stake delegated to it there routes through its children after that.
ValidatorCrucible LabsSubnetsSN8, SN44, SN76, SN81, SN91
Between blocks 9,204,039 and 9,204,113
Reliquary | Bittensor SN81 τ@reliquary_aiA week ago we added a new task to Subnet 81: batch generation. You give it a dataset and a model, miners generate the completions. Since then, the network has generated nearly 3B tokens with Qwen3.8-27B, with outputs spot-checked against the model's weights. We're using them to distill into Teutonic, a model from Bittensor. Next: making this available to everyone. More on Tuesday.
SN81 large buy: 548.4 τ
SN81 large buy: 335.8 τ
SN81 large sell: 708.0 τ
Reliquary | Bittensor SN81 τ@reliquary_aiNew task type coming to Reliquary, alongside RL: SFT data generation. The first job, miners generate reasoning traces with Qwen3.8-27B on math and code, and every single token is verified against the model before it gets paid.
Reliquary | Bittensor SN81 τ@reliquary_aiFor small models, distilling from a much stronger teacher beats RL alone. That's the plan for Teutonic: SFT on a verified corpus from the 27B, then RL on top. Both run on Reliquary. This is the first of several new tasks to build a full post-training layer.
Reliquary | Bittensor SN81 τ@reliquary_aiWe just shipped a major upgrade to Reliquary's validation system. Validators no longer need to load the full model. They stream it layer by layer, so a single H100 can now verify a frontier MoE, bit-identical to running it whole. Bigger models are coming to Reliquary.
Reliquary | Bittensor SN81 τ@reliquary_aiReliquary-4B is a research release and a foundation for what comes next. We’re releasing the weights and a technical report covering the training run, results, and system design. Model: huggingface.co/ReliquaryForge/Reliquary-4B Paper: reliqua.ai/research/rollout-selection-mechanism…
Linkhuggingface.cohuggingface.co/ReliquaryForge/Reliquary-4B