Subnets / SN74
Gittensor SN74
Gittensor (subnet 74) is a Bittensor subnet held by 1,688 unique coldkeys. Its liquidity pool holds 6.1k τ, with the alpha token trading at 0.0034 τ. Roughly 16% of tracked holders are in profit — most bought in above today's price. Seller exhaustion is elevated: 89% of recent sellers have emptied their wallets, thinning the supply overhang. Full insider flow, whale tracking and the complete proprietary factor set for Gittensor are available to SubnetStats subscribers.
ALPHA PRICE
0.0034 τ
IN TAO
HOLDERS
1,688
UNIQUE COLDKEYS
POOL DEPTH
6.1k τ
TAO IN POOL
7D MOMENTUM
-4.4%
ALPHA vs TAO
Project activity
first closed source. open source is next.
if your code can be taken down with one email to github, how permissionless is it?
@const_reborn we need code hosting baked into the chain.
. @sama @elonmusk @DarioAmodei want to decide how fast AI moves and who competes.
bitcoin’s power is that nobody gets to decide who can participate.
if only there were a blockchain where anyone could build and compete to produce intelligence..... https://t.co/d2PuiFmngy
anthropic, openai and xai suddenly agreeing the industry needs to slow down should worry you.
if you don’t see the oligarchy forming, i can’t help you. https://t.co/Eg1Cj0utXo
We disagree. Time to accelerate.
25% faster generation. 43% more room for context. Comparable accuracy.
Gittensor vs @nvidia’s NVFP4 build of Qwen3.8-27B, tested on the same RTX 5090. https://t.co/zblXbAZZyQ
agent civilizations https://t.co/eyshxFk9Tu
262,813 Hugging Face downloads in 17 days for our RTX 5090-optimized Qwen3.8 checkpoint.
We made the most downloaded model release ever from a Bittensor subnet. Developers are running, benchmarking, and building on it.
Gittensor API coming this week.
https://t.co/iwkP0jstmL
just merged another Qwen3.8 speedup on RTX 5090.
Its scored DSpark path at 4k context went from 90.04 to 106.40 tok/s, while keeping lossless target-token equality.
That is an 18.2% gain from one merged update.
59k downloads in 4 days for our RTX 5090-optimized Qwen3.8 model
https://t.co/iwkP0jstmL
39.7k downloads within 3days
New release version,
Qwen3.8-27B-NVFP4-RTX5090
The flagship: NVIDIA ModelOpt NVFP4 of Qwen3.8-27B, tuned for GeForce RTX 5090.
Full 262K context on 32 GB 81.6 tok/s alone — 155.8 tok/s with the drafter
https://t.co/iwkP0jstmL
Metagraphed turned Bittensor into something an agent can actually navigate.
Before this, if you wanted to understand Bittensor you had to manually piece it together from scattered docs, dashboards, repos, APIs, and subnet sites.
This MCP gives an agent one machine-readable entr
Bittensor needed a front door.
Now it has one.
Give this to your agent and it gets a live map of Bittensor: what every subnet does, and exactly how to use it.
Copy
```claude mcp add --transport http metagraphed https://t.co/90Vb4ZiV9G```
Gittensor makes open source profitable
There are only 2 ways to get more inference: add more compute, or get more tokens per second out of the compute you already have
The future of AI is open source.
Gittensor supports the push for open-weight AI models, open standards, interoperability, and real competition across the stack.
We’d be proud to stand alongside @Microsoft @NVIDIA @Google @Meta and the organizations backing this letter: https://
An inference engine for consumer Blackwell GPUs, SparkInfer is now pushing about 17,015 prefill tokens per second at 128k context on a single RTX 5090.
v0.4.3 turned long-context prompt processing into the headline.
That is the kind of speed local agents actually feel.
Prefill is the stage where the model reads the input prompt and builds its internal KV cache before generating any output tokens.
KV cache is the saved internal memory of the prompt’s attention state, so the model doesn’t have to recompute the whole context every time it generat
Choosing a Claude tool is becoming a workflow, not a scavenger hunt.
The latest HeyClaude changes turn category, source, and trust badges into the next browse or compare move, so the directory helps you decide instead of just handing you links. https://t.co/9RaWnyp4i4
Most Claude tooling directories still make you leave the page to judge trust.
HeyClaude is moving the opposite way: source, install risk, safety, and privacy signals now sit in the browsing flow before the install step.
Source and product flow:
https://t.co/ecy09LHp3z https://t
Most Claude tooling directories still stop at the link.
HeyClaude is building the missing inspection layer before install: a reviewed directory for Claude tooling where builders can browse, compare, and check trust signals before copying anything into their workflow.
The live r
https://t.co/NDYvSFTtk2
Most self-hosted AI review tooling still makes you assemble the stack yourself.
LoopOver, formerly Gittensory, just shipped Orb v0.6.0-beta.1: a published multi-arch image for its review stack with Claude Code and Codex CLIs already in the container, while credentials stay runti
Codex and Claude are drugs, their goal is to get everyone addicted
local consumer gpu inference is moving faster than most people realize.
sparkinfer now leads its llama.cpp baseline on 3 different qwen models on the same rtx 5090:
qwen3-moe: 480.7 tok/s
qwen3.6: 424.9 tok/s
qwen3.5: 279.8 tok/s
same box. same decode test. three wins. htt
SELLERS DRAINED
89%
28D SELLERS EMPTIED OUT
CAPITULATION
-0.44
BOUNCE COMPOSITE (z)
HEAT 7D
0.12×
VOLUME ÷ POOL DEPTH
HOLDERS IN PROFIT
16%
TRACKED WALLETS · ENTRY BELOW PRICE
See who's buying and selling Gittensor — every trade classified miner / validator / owner / outsider, plus holder cost basis and the full 23-factor screener.
OPEN THE INSIDER TAPE →