Bankr Founder Trains On-Device Model on Anonymized Agent Sessions
@0xDeployer said he is training a small Bankr model on tens of thousands of anonymized BankrBot agent sessions to keep conversations on the user's phone.
Also in: 日本語 · エージェント経済 · 한국어 · 에이전트포스트 · 简体中文 · 链智日报
@0xDeployer, the founder of Bankr, said on October 5, 2026 that he is experimenting with a small Bankr model intended to run on a user’s own device, and that he is training it on tens of thousands of real, anonymized BankrBot agent sessions.
In a post on X at 00:03 UTC on October 5, 2026, he described the aim as keeping a user’s conversation with their agent on their phone, with no cloud LLM reading their trades. He did not name the model, give its parameter count, or say which devices it would run on. He said he would share how it performs.
BankrBot is a trading agent framework built on Bankr, a platform that pays token creators a share of trading fees and provides funding for launches. Bankr’s documentation states that 95% of a 0.7% pool swap fee, or 0.665% of trading volume, goes to the token creator and is claimable at any time. The platform vests 15% of supply to creators over one year with a 30-day cliff and enforces a 2% wallet cap for the first five minutes after launch.
The training data described in the post would come from agent sessions, the conversations and actions recorded when a BankrBot agent runs. The post says the sessions are anonymized and does not describe the anonymization method, the retention period, or where the training runs.
The disclosure follows two other Bankr experiments this month. On October 2, 2026, @0xDeployer began a 10-day trading challenge in which a BankrBot agent managed trades without human intervention, and he said no part of that process was managed by a human. The agent posted its first take-profit hit the same day, after a dry run that recorded 75% wins, according to his post at 21:01 UTC on October 2, 2026. The live trade opened at base address 0xfb31f85a8367210b2e4ed2360d2da9dc2d2ccc95, and the agent was trailing the remaining position. The post did not disclose the initial trade size, the asset traded, or performance beyond the dry-run win rate.
On October 1, 2026, he said almost 25% of the BNKR token supply was staked, up from 15.4% on September 27, 2026. Bankr opened early staking on September 26, 2026, the same day he wrote that the company had bought back seven figures of BNKR to stream to stakers. Bankr’s staking documentation states that rewards are paid in BNKR and streamed weekly, with each week’s batch released second by second over the following seven days, and that a staker’s share of the pool depends on total weight staked.
The October 5 post did not say when the model would be released, whether it would be open to users outside Bankr, or how it would be evaluated.