The verification and settlement layer for global AI inference.
Mainnet
Joined June 2026
- Tweets373
- Following8
- Followers1.5K
- Likes102
Pinned Tweet
A new look for TensorCash.
The website and documentation now explain the whole protocol, from proof of inference to the financial layer.
Welcome to the verification and settlement layer for global AI inference.
tensorcash.org/?v=2
TensorCash retweeted
On TensorCash, inference is mining.
Qwen3.8-27B, served live by a real miner's GPU. The miner's name sits under every reply.
That same work secures the chain, so anyone can prove which model actually answered.
Just the beginning 👀 @Tensorcash
TensorCash retweeted
The thesis inferenceFI was founded on: inference will trade like any other commodity.
Two models can now be mined on TensorCash: Qwen3-8B and Qwen3.8-27B.
The newly added 27B has more than three times the parameters, making it more capable. That also means each 256-token window takes more compute, so a miner would get fewer attempts at a block.
To make up for that, and keep both models worth mining from a block reward perspective, its difficulty is set lower. A miner’s chance at a block stays roughly the same on either model.
ALT Comparison card: Qwen3-8B, a lighter model with more attempts per hour and a higher difficulty, next to Qwen3.8-27B, with about 3x the parameters, fewer attempts per hour and a lower difficulty.
Both models are being mined right now.
Check it yourself:
Qwen3-8B
scan.tensorcash.org/model/4b…
Qwen3.8-27B
scan.tensorcash.org/model/c5…
Questions have come in about community-built web wallets, extensions and EVM bridges.
No official EVM bridge for TensorCash has been released. A supported bridge will only be announced here and on tensorcash.org.
Until then, no bridge is official.
Several community wallets have been reviewed, and their integrity checks are published: integrity.tensorcash.org/ver…
TensorCash is a proof-of-work chain where the work is AI inference. Miners can sell this inference and get a chance at the next block with the same work.
It sets the standard for verifiable inference by attaching a proof of the model used to every answer.
Bob Labs runs a TensorCash mining pool, plus a chat app and an API where users and developers buy inference from its miners.
Usage is prepaid in $TSC and passed on to the miners who serve the requests.
One day after Qwen3.8-27B went live on mainnet, it is already serving paid traffic: a full coding session, served by a TensorCash miner and sold through @AskTillyAI, an independent operator on the network.
Every request came with a proof of which model did the work.
An autonomous coding agent ran for 28 minutes on a @Tensorcash miner via the AskTilly API. One prompt, and it built this animation on its own for the client.
We sold the session for $0.16 and the miner was paid in tensorcash:native. A day after the model went live, already at market rates.
The miner was paid in $TSC, and because the same work also earns block rewards, miners can compete on price in a way standalone providers cannot.
Miner payment: scan.tensorcash.org/tx/e408d…
TensorCash retweeted
Qwen3.8-27B got pushed live on @Tensorcash mainnet today, with benchmarks putting it up there with Opus 4.6 on coding
Saw this interesting data:
~231M Qwen downloads on HF in 30 days > even if part of that is data centres pulling the same weights over and over, it’s still 7x the next model developer, showing real adoption.
Labs should let networks like TensorCash serve their models and keep their own GPUs for training the next ones.
ALT Bar chart of Hugging Face downloads by model developer, last 30 days: Qwen 230.8M, Google 30.6M, Nvidia 25.4M, DeepSeek 25.0M, Meta 18.6M.
Qwen3.8-27B (bf16) is now registered on TensorCash mainnet, a model reported as competitive with today’s top models.
From now on, coding and autonomous agent tasks can be bought as inference from the TensorCash network, a large step up from Qwen3-8B.
Block producers and miners are advised to upgrade to 1.2.3a-r2 before mining with the new model.
Qwen3.8-27B (bf16) is now registered on TensorCash mainnet, a model reported as competitive with today’s top models.
From now on, coding and autonomous agent tasks can be bought as inference from the TensorCash network, a large step up from Qwen3-8B.
Block producers and miners are advised to upgrade to 1.2.3a-r2 before mining with the new model.
Check it for yourself: scan.tensorcash.org/models
Release notes: git.tensorcash.org/tensorcas…
TensorCash 1.2.3 activates today at block 31,400, expected around 13:00 to 15:00 UTC.
Verifiers: upgrade to 1.2.3a-r1. Full wallets: upgrade from the release page.
After activation, a registration for Qwen3.8-27B (bf16) is planned, with a difficulty multiplier of 50,000.
Full wallet: git.tensorcash.org/tensorcas…
Verifier 1.2.3a-r1: git.tensorcash.org/tensorcas…
An early contributor of the protocol signed his first post with a qualified electronic signature, and the proof is stored on TensorCash.
The hash of the text, the certificate fingerprint and the signature are in block 31,144. This shows that the exact text was signed and existed at that block, and it cannot be changed or backdated. The certificate ties the signing key to his legal identity.
Native assets on TensorCash offer additional rules that every node enforces. Among others, the signed legal document belongs to the asset, no more units can exist than it allows, and it can only change with the holders’ approval.
Some questions were raised about this account not being real, given its long inactivity, so here is the cryptographic evidence. The hash of this tweet and a digital signature of the very same are published here, as they say in @Tensorcash , verify it yourself: asktilly.ai/verify/enrico-ot…
Check it for yourself: asktilly.ai/verify/enrico-ot…
How it works for native assets: tensorcash.org/blog/rwa-the-…