Subnets / SN25
Mainframe SN25
Mainframe (subnet 25) is a Bittensor subnet held by 1,569 unique coldkeys. Its liquidity pool holds 10.5k τ, with the alpha token trading at 0.0041 τ. Roughly 3% of tracked holders are in profit — most bought in above today's price. Seller exhaustion is elevated: 87% of recent sellers have emptied their wallets, thinning the supply overhang. Full insider flow, whale tracking and the complete proprietary factor set for Mainframe are available to SubnetStats subscribers.
ALPHA PRICE
0.0041 τ
IN TAO
HOLDERS
1,569
UNIQUE COLDKEYS
POOL DEPTH
10.5k τ
TAO IN POOL
7D MOMENTUM
-4.8%
ALPHA vs TAO
Project activity
first 24h of iota-decide-4b:
@huggingface built a free demo space for it
@pratikg benchmarked it on a mac mini at ~90ms a decision
We’re testing it internally as an agentic control plane for @IOTA_SN9, so the swarm can diagnose and recover from its own failures in real time.
Introducing IOTA Decide 4B, our first public model.
Pareto optimal. Built different.
Trained across globally distributed GPUs with our SDK.
Open weights, Apache-2.0.
🤗 https://t.co/cFlqLmcnop
The world doesn't have a shortage of compute. It has a problem connecting it.
We've built the infrastructure to change that.
The future is liquid.
Hundreds of worker failures and the result is the same.
We don't think about fault tolerance any more.
Never stop training.
Day 1 at #GTCBerlin. Neoclouds are racing to optimize inference and turn raw compute into turnkey training products. The Overton window on distributed training is shifting, and we are ready. https://t.co/tnq0hp0v7I
Train anywhere, train everywhere.
Apex <> Bitstarter launch https://t.co/QKIoEWLzHR
In a world where answers are cheap, the right questions are everything.
Welcome to the largest optimization landscape in the world @bitstarterAI. We think you'll like it here.
One week since we announced the IOTA SDK at @ExploitSummit . What our team have done with it so far:
- Trained a 16B across 139 consumer GPUs ( 4090/5090), with 1.8x wall time speedup vs our legacy codebase. Butterfly all-reduce times reduced by 30% because of powerful new abstr
If you run a fleet, idle capacity is revenue you're not earning.
Pretraining is only one job. The iota SDK also runs RL and post-training, fine-tuning, and inference on the same disaggregated compute. https://t.co/WuQbKqOSgG
The way teams get training compute today holds them back. That's why we're taking iota to market. https://t.co/6USEBPm1il
The iota SDK makes fragmented, interruptible compute train like one centralised cluster. Here's how, in two parts: the engine and the steering wheel. https://t.co/Y8kyvu3q72
Read our CEO @WSquires' thoughts from @ExploitSummit, where we announced the release of the iota SDK this year.
The SDK allows anyone to harness distributed, heterogenous, and idle compute to create frontier scale training outcomes.
https://t.co/WsnRgksELF
Orion is how we test whether the long tail can train serious models. Two runs so far, at 16B and 100B parameters. Here's what they showed.
Demand for AI compute is growing far faster than the infrastructure being built to serve it. Sam Altman's stated goal for OpenAI is 250 GW by 2033, roughly a quarter of US generating capacity. If the trend holds, Epoch AI puts a single frontier training run at 4 to 16 GW. https:/
Scattered compute can now train like one cluster.
Yesterday at @ExploitSummit we announced the iota SDK and Liquid Compute, our disaggregated compute platform. Here's what they unlock. https://t.co/oq779UTZQK
Just announced earlier today at @ExploitSummit in Montreal: the iota SDK and Liquid Compute.
Liquid Compute is our disaggregated compute platform. It turns the long tail of global compute into capacity you can actually train on.
The iota SDK powers any training workload across
Just announced on stage at @ExploitSummit in Montreal: the iota SDK and Liquid Compute.
Liquid Compute is our disaggregated compute platform. It turns the long tail of global compute into capacity you can actually train on.
The iota SDK powers any training workload across it, a
Live now at @ExploitSummit the Macrocosmos co-founders @macrocrux and @WSquires are on stage presenting iota now!
Distributed training across scattered hardware, coordinated as one system.
Watch live at the link below.
https://t.co/wnETRODYw6
Training a 100B parameter model normally means renting a block of matched, high-spec hardware and holding it unbroken for the length of the run.
Orion-100B ran across five data centres instead, on single A100s at around $1.25 an hour each. That puts a full replica at roughly $20
Exploit Summit opens Monday in Montreal.
We'll be talking about training on compute that was never set aside for the job, and what it takes to keep a run alive when the hardware underneath it keeps changing.
Find us there.
We trained Orion-16B without a reserved cluster anywhere in it.
The numbers underneath that:
It sustained around 80k tokens per second across 4090s, 5090s, A100s, L40s, and A6000s, at roughly 20% MFU. The accessible pool came to about 18 B200s equivalent at FP16.
Nodes dropped,
We're at Exploit Summit in Montreal next Monday and Tuesday.
On stage: what it actually takes to train a model on compute nobody set aside for you. Mixed hardware, several continents, machines joining and dropping mid-run, one finished model at the end of it.
Come and say hello
SELLERS DRAINED
87%
28D SELLERS EMPTIED OUT
CAPITULATION
+0.20
BOUNCE COMPOSITE (z)
HEAT 7D
0.25×
VOLUME ÷ POOL DEPTH
HOLDERS IN PROFIT
3%
TRACKED WALLETS · ENTRY BELOW PRICE
WHAT'S CARRYING THE PRICE · 30DMIXEDemission rank #51 · below the bar (rank 32)
30D RETURN
-37.8%
ALPHA vs TAO
WITHOUT CHAIN BUYS
-38.3%
SAME 30 DAYS, PROTOCOL BID REMOVED
CHAIN SHARE OF BUYING
100%
CHAIN BUYS vs TRADERS' NET
CHAIN BUYS / DAY
0 τ
PROTOCOL BUYING ITS ALPHA
See who's buying and selling Mainframe — every trade classified miner / validator / owner / outsider, plus holder cost basis and the full 23-factor screener.
OPEN THE INSIDER TAPE →