Subnets / SN114
SOMA SN114
SOMA (subnet 114) is a Bittensor subnet held by 2,227 unique coldkeys. Its liquidity pool holds 6.9k τ, with the alpha token trading at 0.0146 τ. Roughly 92% of tracked holders are in profit — the holder base is broadly in the green. Seller exhaustion is elevated: 78% of recent sellers have emptied their wallets, thinning the supply overhang. Full insider flow, whale tracking and the complete proprietary factor set for SOMA are available to SubnetStats subscribers.
ALPHA PRICE
0.0146 τ
IN TAO
HOLDERS
2,227
UNIQUE COLDKEYS
POOL DEPTH
6.9k τ
TAO IN POOL
7D MOMENTUM
+4.1%
ALPHA vs TAO
Project activity
39.7% cheaper input with SOMA.
We tested our context compressor against GPT-6 Astra.
The original prompt used approximately 12,937 tokens, with an estimated input cost of $0.1294.
We then took the exact same prompt, compressed it with SOMARIZER at a 0.60 compression ratio, and
SOMA Highlights - September 9
We opened SOMA to everyone. One competition later, savings went from 10% to up to 15%.
Try it in GitHub Copilot with DeepSeek V4 Pro.
This time @japanese_crispy joins @oli_soma.
Watch this week’s update below.
Compress. Pay less. https://t.co/fHM
We taught SOMA a new trick.
The new compression algorithm improved on the previous one by 5pp, so you can now save up to 15%.
Put it to work: https://t.co/CF6bySeyXn
Compress. Pay less. https://t.co/6wXbSYKaQi
SOMA sits between the agent and the model.
Prompt → Agent → SOMA → LLM
It compresses the task list before it reaches the LLM, so the model reads fewer input tokens and still solves the same task.
Compress. Pay less. https://t.co/00qqF43UZF
Tokens are units of time.
For AI agents, context compression turns those tokens into more working time.
More work from the same compute.
SOMA is now open.
Plug it into GitHub Copilot and run DeepSeek V4 Pro with context compression.
Try it on a real coding workload.
Tell us what works, what breaks, and what we should improve.
More models coming soon.
Try SOMA: https://t.co/etBanUYNgi https://t.co/I4Qfxj3Ks2
SOMA is live for GitHub Copilot.
All Early Access emails have now been sent. If you signed up, check your inbox.
You can plug SOMA into Copilot and use DeepSeek V4 Pro with 10% savings. Every Early Access user gets $5 in credits to test it on real coding tasks.
There are no SO
SOMA Highlights - August 30
➡️ SOMA Early Access product coming soon
➡️ @say_gm_ integrated the SOMA algorithm
➡️ Introducing the concept of hybrid competition
Full breakdown below 👇 https://t.co/fmj4vG3i5h
We’re bringing SOMA to GitHub Copilot on August 31.
➡️ Around 10% token savings
➡️ Payments in fiat or $TAO
➡️ $5 in credits for early users
Over time, we’ll expand the product with more models, more harnesses, and greater savings.
Sign up now: https://t.co/F1S7McihVL https://
The best product feedback is a production deployment you didn’t ask for.
@say_gm_ ported SOMA’s OpenClaw compression core to Rust and is running it inside their privacy-first gateway for dozens of AI models.
That’s what we built SOMA for. https://t.co/gMS8eJ9LVF
This is the SOMA team - we're definitely real.
Engineers and researchers who have spent years on variuos projects, distributed systems and incentive designs, backed by @DendriteHQ.
@oli_soma - CEO of SOMA
Matt - Owns SOMA's technology end to end and leads the engineering org:
Agentic workloads are turning context into one of the largest components of inference spend.
As agents run longer loops and make more calls, the same context gets processed again and again.
Join to SOMA Early Access here: https://t.co/F1S7McihVL
Agentic inference changes the economics of tokens.
A chat request consumes context once. An agent can reuse and expand the same state across dozens or hundreds of inference steps - which makes context size, KV-cache pressure and memory bandwidth increasingly important bottleneck
SOMA Highlights - August 21
➡️ Early access sign-ups are open
➡️ @DendriteHQ team is testing the SOMA compressor
➡️ 10% token savings achieved this round - 20% is the next target
➡️ Scout, our coding task generation pipeline, is in final testing
➡️ Compression models are now und
Live now: @oli_soma with @gordonfrayne talking SOMA - context compression and more insights into what we're building on SN114.
Everyone building on language models pays per token.
Few can explain what a token is, how it gets counted, or why input tokens dominate agent workloads.
We wrote it out from the bottom up, tokenizer to invoice:
https://t.co/VRz43ayScx
39,9% cheaper input with SOMA.
We tested our context compressor against ChatGPT 5.5 Pro, using a simple prompt for a 2D HTML/CSS animation.
The original prompt used approximately 8,680 tokens, with an estimated input cost of $0.5729 (at $10/MTok), and contained 31,364 character
Powerful enough to be banned, still only 11% of business spend. The reason given is price.
In agent workloads, much of that cost comes from input tokens, often context the model never needed to see.
We compress that context before inference. Same model, smaller bill.
🚨 New SOMA Tool is coming.
Copy a command, connect it to your agent, and keep using your existing workflow as usual.
SOMA runs as an optimization layer that saves you up to 50% on AI costs.
Same quality. Lower cost. https://t.co/zbQ8sveUdN
SOMA Roadmap - where we actually are.
[Shipped ✅] Plain-text compression powers SOMARIZER. Set the level - the tighter the ratio, the bigger the token savings.
[Shipped ✅] CoT compression for OpenClaw. The compressor processes the full session state, including user messages, re
We are postponing the launch of the SOMA SN114 Conviction Program. 🧵 1/2
The program uses Bittensor’s conviction system. In simple terms, locking subnet tokens behind a key gradually increases the strength of that key. If another key gains enough strength, it can replace the cur
The repository is live.
SOMA’s context compression plugin for @openclaw is now open source.
Plug it into your existing pipeline to reduce token usage without sacrificing performance.
Link below 👇 https://t.co/XPqPZIRemT
Agents rarely fail because the model is weak. They fail because context is bloated, expensive, and mostly wasted.
Every oversized prompt, every long reasoning chain, every retry is spend you already committed to.
So we built the fix, and we are opening it up.
SOMA ships its fi
https://t.co/aGyiZlHx9F
SELLERS DRAINED
78%
28D SELLERS EMPTIED OUT
CAPITULATION
-0.72
BOUNCE COMPOSITE (z)
HEAT 7D
0.98×
VOLUME ÷ POOL DEPTH
HOLDERS IN PROFIT
92%
TRACKED WALLETS · ENTRY BELOW PRICE
See who's buying and selling SOMA — every trade classified miner / validator / owner / outsider, plus holder cost basis and the full 23-factor screener.
OPEN THE INSIDER TAPE →