Subnets / SN28
gm SN28
gm (subnet 28) is a Bittensor subnet held by 2,634 unique coldkeys. Its liquidity pool holds 23.1k τ, with the alpha token trading at 0.0228 τ. Roughly 95% of tracked holders are in profit — the holder base is broadly in the green. Over the last 7 days the alpha price is up 0.1% in TAO terms. Full insider flow, whale tracking and the complete proprietary factor set for gm are available to SubnetStats subscribers.
ALPHA PRICE
0.0228 τ
IN TAO
HOLDERS
2,634
UNIQUE COLDKEYS
POOL DEPTH
23.1k τ
TAO IN POOL
7D MOMENTUM
+0.1%
ALPHA vs TAO
Project activity
gm 47 models and 31% average discount on list price. https://t.co/DpncDHuoT7
Hammer away 🔨
This is what SayGm was built for: bring privacy and savings to the tools people already use 🤝
Your AI provider says it won't read your prompts. A TEE means it can't.
We wrote up how TEE's work, why the prompt is often sensitive (not just the output), and how to check a provider's privacy claim instead of taking their word for it.
Read here 👇
https://t.co/n77zyIUkMD
Deepseek-v4.1-flash is live on SayGm.
Natively multimodal, 1m token context, and deepseek's own benchmarks put it ahead of v4-pro on performance, cost, and speed.
And cheaper than list pricing, as always: 5% off.
You can soon run @webuildscore image generation, training, and model deployment through SayGm's inference routing.
Score Studio launching on Sept 16. https://t.co/izW0nCw8oE
In recent AI news: a mathematician asked OpenAI whether his unpublished sessions trained the model behind its Navier-Stokes result.
The answer covered access, not training, two separate questions.
We break down what that means for your own work 👇
https://t.co/QNu50jSUKv
GM is approaching 100 billion recorded network tokens processed.
We're at 96.48B.
Including more than 30B output tokens generated. 🧵 https://t.co/jYLRRn6VcO
GPT-6 Astra launched. The White House shipped an arcade. Same week.
So we gave Claude Fable 5.1 and GPT-6 Astra one prompt each: build a self-playing arcade game for a G7 country.
No retries, names hidden until the reveal.
Six games. Can you tell which one is GPT-6 before
GPT-6 Astra is live on Gm.
OpenAI’s smartest model yet. Tell it what you need done. It opens the browser, writes the code, runs the terminal, ships the result.
Routed through a verifiable TEE, at 8.3% discount on SayGm.
https://t.co/sk0GHhAmTh
A shared dependency. It happens.
Two answers from gm: every model runs across several providers, so a single route failing doesn't take you down. And you can define a cascade, your model order, your rules, and the gateway falls through it automatically.
Point your existing SDK
Gemini's newest model: Gemini 3.8 Flash is live on SayGm now.
8.3% cheaper than list pricing.
Also shipped: image generation support for Gemini models through the API.
Be the first to try it: https://t.co/Wb4qKHXf8v https://t.co/3ROcl6lTAI
Fable 5-1 costs more per call than Fable 5.
It's still the cheaper option, because one rocket crashes and the other sticks the landing 🚀
Both models available now on SayGm. https://t.co/L5oIyBPpBK
If you're paying list pricing for Fable 5, just know you could be paying 64% less on gm ☀️
Claude Fable 5.1 from Anthropic is live on Gm with an 8.1% discount.
Fable 5.1 raises the ceiling on what Fable 5 could do, particularly for coding agents, multi-step workflows, generating code from visual inputs, and financial or analytical work.
Try it now 👇
https://t.co/z3y6
DeepSeek started the AI price war. Cut inference costs by over 200x in three years. Every dev built their stack assuming tokens only get cheaper.
This month, DeepSeek told developers it's raising prices. No new rate card yet. Just a warning that it "could be significant."
Here'
GLM-5.3 weights just dropped - and it’s already live on SayGm.
Another day-one launch, served by @KubeTEEAI.
GLM-5.3 vs GLM-5.3-Flash.
Can you guess which is which before the reveal? https://t.co/rOXupGlCJo
Your SayGm tool calls just got lighter.
Test logs, json dumps, file reads. The bulky stuff that used to eat your agent's context window now gets compressed first, and it's done the smart way. Every line your agent actually needs stays untouched. Everything else gets cut.
No ext
New models, day one showdown.
Opus 5 and GPT 5.6 Sol vs GLM-5.3 Flash and Qwen3.8 Flash Next, the last two released yesterday.
Btw if you've been running "Ox Alpha" in stealth, that's GLM-5.3 Flash. Both live on SayGm now.
Can you guess which model is which before the reveal?
58% cheaper on @say_gm_ 👇
https://t.co/PDIzO4eIkD
New models just dropped and they're already live on SayGm 🫡
→ GLM-5.3-Flash — 58.8% below @Zai_org list price
→ Qwen3.8-Flash-Next — 12.4% below @Alibaba_Qwen list price
Both models are less than 24 hours old.
Be one of the first to try, courtesy of @KubeTEEAI. https://t.co/8R
DeepSeek V4 Flash is 75% cheaper on saygm than list price.
Not a promo code. Not a limited window. Live provider bidding, updated in real time.
Openrouter shows you the sticker price. We show you what it actually costs. https://t.co/PngVw9JbLm
Hey @ornith_ Ornith-1.5-397B has been live on SayGm since 20 Aug.
We're the first hosted API for it that we know of. Solid model, congrats on the HN front page.
For anyone who wants to try it without standing up 8x80GB GPUs: https://t.co/f4NVke6Mzj https://t.co/y93GXC1LQI
Testing something new: gm showdowns
Same prompt. Opus goes for complexity, gpt-5.6-sol gets close for a quarter the price.
Both cheaper on SayGm than retail too: opus -34%, gpt -54%
No winner called here, you tell us?
Ornith-1.5-397B just topped Hacker News.
It’s live on SayGm right now, and we’re the first platform serving it.
It matched Claude Opus 4.8 on every answer in our reasoning tests, and its output tokens cost roughly a quarter as much.
Be first to try 👇
https://t.co/YMTLwpuiiD ht
SELLERS DRAINED
66%
28D SELLERS EMPTIED OUT
CAPITULATION
-1.36
BOUNCE COMPOSITE (z)
HEAT 7D
1.17×
VOLUME ÷ POOL DEPTH
HOLDERS IN PROFIT
95%
TRACKED WALLETS · ENTRY BELOW PRICE
See who's buying and selling gm — every trade classified miner / validator / owner / outsider, plus holder cost basis and the full 23-factor screener.
OPEN THE INSIDER TAPE →