Skip to content
The BCI Briefing

Technology4 min read

Neuralink says pooled participant data pushed cursor control to 11.32 bits per second

The company reports that decoders pretrained on more than 50,000 hours of neural recordings set a new internal record and cut weekly recalibration. The results are not peer reviewed.

ByThe BCI Briefing

Evidence class: Reported claim

The short version

  • One participant reached 11.32 bits per second; the company's previous best was 10.39.
  • Decoders were pretrained on more than 50,000 hours of unlabelled recordings from study participants.
  • Some participants reportedly cut recalibration from about 55 to about 10 minutes a week.

In a blog post dated 1 October, Neuralink described self-supervised neural encoders pretrained on more than 50,000 hours of freeform neural data streamed by its participants over roughly two years. The first participant alone is said to have contributed more than 9,000 hours.

According to secondary reporting of the post, one participant set a cursor-control rate of 11.32 bits per second, three participants exceeded the previous record of 10.39, and the median across participants was around 10.

The company also reported longer periods of use between recalibrations: one participant maintained control for five consecutive days and another for more than three weeks.

How solid is this

Company reported. We were unable to retrieve the original post automatically and rely on secondary reporting of it. Bits per second depends heavily on the task and how it is scored; figures from different groups are not directly comparable. No peer-reviewed clinical results from the PRIME study have been published.

Company reported. Stated by the company itself in a release, blog post or presentation. Not independently verified and usually without full methods.

What to watch

  1. 01A peer-reviewed description of the method and the evaluation task.
  2. 02Independent benchmarks for decoder stability, such as those now public for academic datasets.
  3. 03Evidence that higher bit rates translate into greater independence in daily activities.

This technology remains experimental and is not an established treatment. Access is limited to registered clinical studies.

Sources

  1. 01

    Neuralink Sets New BCI Record After Training on 50,000 Hours of Brain Data (opens in a new tab)

    Independent reportingTesla North2 Oct 2026Checked against source 3 Oct 2026

    Secondary report of a Neuralink blog post dated 1 October 2026.

  2. 02

    Neuralink — Updates (opens in a new tab)

    Company reportedNeuralink1 Oct 2026Source identified, not opened

    Primary company post could not be retrieved automatically; figures are taken from secondary reporting of it.

  3. 03

    ClinicalTrials.gov NCT06429735 (opens in a new tab)

    Clinical registryClinicalTrials.gov9 Jan 2026Checked against source 3 Oct 2026

    Registry record last updated 2026-01-09.

  4. 04

    Kaifosh P et al. Nature 2025;645(8081):702–711 (opens in a new tab)

    Peer reviewedNature23 Jul 2025Checked against source 3 Oct 2026

    DOI 10.1038/s41586-025-09255-w

  5. 05

    github.com/neuro-galaxy/torch_brain (opens in a new tab)

    Open-source repositoryGitHubChecked against source 3 Oct 2026

    Repository metadata read from the GitHub API.