What Is Neuralink’s New BCI Speed Record? Explained

Neuralink says one of its brain-implant trial participants hit 11.32 bits per second (BPS) — a new high for the company — after training its decoding software on more than 50,000 hours of recorded brain activity. That’s Neuralink’s new BCI speed record, and in practice it means people using the implant can move a cursor or type faster, with far less daily setup, than they could a year ago. Here’s what the record actually measures, how Neuralink says it got there, and why it matters even if you’ll never have a chip implanted.

Key Takeaways

• The Neuralink BCI reached 11.32 bits per second (BPS) in one trial participant, a new personal record, up from its prior 10.39 BPS benchmark.
• The gain came from “pretraining”: the company trained each participant’s neural decoder on tens of thousands of hours of that person’s own past brain recordings.
• Some participants cut their weekly calibration time from an average of 55 minutes down to roughly 10 minutes.
• Some decoders kept working well for three or more weeks without retraining, and one participant has used essentially the same underlying model for over a year.
• Neuralink’s PRIME trial has now reached dozens of participants, and Elon Musk says the company plans to move to high-volume, largely automated implant surgery in 2026.

What the New Neuralink BCI Record Actually Announced

On October 1, 2026, Neuralink said it had started putting more than 50,000 hours of neural recordings to work — data collected from clinical trial participants over the roughly two years since its human trial began. Rather than treating each new task as a blank slate, the company is now using each participant’s accumulated recording history to train that person’s own decoding model before it’s ever used for something like cursor control.

The headline number is 11.32 bits per second, set by one participant and surpassing Neuralink’s previous best of 10.39 BPS. The company says the median across participants is now around 10 BPS, and an experimental “cursor teleportation” system — where the cursor jumps straight to a target instead of moving toward it step by step — also crossed the 10 BPS mark in testing. (Source)

What Does “Bits Per Second” Mean for a Brain Implant?

Bits per second is a way of measuring how much information someone can communicate through an interface each second, borrowed from information theory. For a brain implant, it’s roughly the brain-computer equivalent of typing speed: the higher the number, the faster a person can move a cursor, select letters, or otherwise turn intention into action on a screen, using only their thoughts.

It’s also the metric Neuralink and its competitors use to track progress over time, because it captures both speed and accuracy in a single number — a system that’s fast but makes constant mistakes doesn’t actually score well, since errors have to be corrected.

How “Pretraining” Made the Jump Possible

Brain signals are noisy, and they drift slightly from day to day and even session to session. Historically, that’s meant recalibrating a BCI’s decoder frequently just to keep it accurate. Neuralink’s fix is to pretrain a participant-specific encoder on thousands of hours of that person’s own past recordings, teaching the system to recognize their particular, recurring patterns of brain activity before asking it to do anything useful. Those patterns get converted into stable internal representations — the company calls them embeddings — that hold up better over time than raw, noisy signal.

It’s a bit like a translator who has spent years listening to one specific person talk, versus a translator meeting them for the first time: the same words come through faster and more accurately, because the listener already knows that person’s quirks. The practical payoff shows up in calibration time, which Neuralink says dropped from an average of 55 minutes a week to roughly 10 minutes for some participants, and in durability — some decoders kept performing well for three or more weeks without being retrained, and one participant has been using essentially the same model for over a year and a half.

How a Neuralink BCI Stacks Up Against Earlier Brain Implants

Neuralink isn’t the first group chasing faster brain-to-computer communication, and the pretraining approach behind its new BCI record builds on decades of academic work. The BrainGate consortium, a long-running research collaboration based at Brown University, has tested implanted BCIs in paralyzed participants since the mid-2000s. In 2026, BrainGate researchers reported two participants reaching roughly 22 words per minute by decoding attempted handwriting rather than cursor movement — a different approach and a different metric than bits per second, but part of the same broader push to turn brain activity into fast, reliable communication. (Source)

Neuralink’s own first human participant, Noland Arbaugh, has become one of the most visible faces of this Neuralink BCI effort. Implanted in January 2024 after a diving accident left him paralyzed below the shoulders, Arbaugh has described the device as giving him a renewed sense of independence, using it to play video games, browse the web, and communicate without assistance. More than 21 months into using the implant, he’s now spoken publicly at technology and robotics events about day-to-day life with a Neuralink BCI. (Source)

The contrast matters: BrainGate’s systems typically rely on external hardware wired to a port on the user’s skull, while Neuralink’s implant is fully internal and wireless. Academic programs like BrainGate have prioritized rigorous, peer-reviewed validation across multiple device generations, while Neuralink has moved faster on hardware iteration and participant count. Both approaches are still feeding the same goal — a Neuralink BCI, or any BCI, that’s fast and reliable enough for daily use outside a lab.

Why This Matters Beyond the Record

Neuralink isn’t working in a vacuum. Competing brain-computer interface companies, including Synchron and Precision Neuroscience, are running their own human trials and raising significant money — Precision Neuroscience alone brought in $250 million this year for its BCI work. Clinical BCI trials broadly have picked up pace through 2026, with multiple companies now racing toward the same kinds of speed and reliability benchmarks Neuralink is reporting here.

The two named competitors take noticeably different hardware approaches than Neuralink’s own Neuralink BCI design. Synchron’s device is delivered through blood vessels rather than open brain surgery, reducing surgical risk at the cost of a less direct connection to neurons. Precision Neuroscience uses a thin, flexible array of electrodes that sits on the brain’s surface without penetrating it, aiming for an implant that can be removed more easily than Neuralink’s threaded electrodes. Each trade-off between signal quality, surgical risk, and reversibility represents a different bet about what kind of brain-computer interface wins out as the field matures.

It’s also worth noticing where this particular gain came from: not a new chip or more electrodes, but a better-trained software model squeezing more signal out of the same hardware. That mirrors the broader pace of AI progress in 2026, where a lot of the most meaningful gains have come from smarter models and more training data rather than entirely new hardware. The same AI techniques reshaping chatbots and AI agents are increasingly what’s driving progress in brain implants, too.

What This Means for the Future of Brain Implants

Elon Musk has said Neuralink plans to start high-volume production of its devices in 2026 and move to a “streamlined, almost entirely automated surgical procedure.” Part of that involves a change to how the implant’s threads are placed: they would go through the brain’s protective membrane, the dura, without needing to remove it first, which Musk has called a significant simplification of the surgery itself. (Source)

Separately, Neuralink has talked about eventually offering implants to otherwise healthy people, not just those with paralysis, though the company has described that as still several years out. For now, every one of these speed gains is happening within a small clinical trial of paralyzed participants, which is worth keeping in mind whenever a brain-implant headline sounds bigger than the current reality.

What a Neuralink BCI Could Mean for Healthy Users

Right now, every working Neuralink BCI is implanted in someone with paralysis, under close medical supervision, as part of a clinical trial. But the company has repeatedly floated a longer-term goal of offering a Neuralink BCI to people with no disability at all — treating it as a new category of consumer technology rather than strictly a medical device. Musk has talked about uses ranging from faster typing and gaming to, more speculatively, restoring sight or hearing.

None of that is close to happening. Even the medical version of a Neuralink BCI is still working through a small clinical trial, under a device designation the U.S. Food and Drug Administration hasn’t cleared for the general market. Brain surgery also isn’t something most healthy people would accept for a speed boost on tasks they can already do with a keyboard. Still, the gap between “works in a lab” and “works well enough to consider outside one” is exactly what metrics like bits per second are meant to track, and Neuralink’s 11.32 BPS record is a sign that gap is narrowing, even if slowly.

What to Watch Next for Neuralink BCI Development

A few concrete milestones will show whether this Neuralink BCI progress holds up. The first is surgical volume: Musk’s 2026 target of high-volume, largely automated implant surgery would mean dozens or hundreds of new participants rather than the handful enrolled so far, and that kind of scale-up tends to surface problems that don’t show up in a small trial. The second is durability outside the lab: Neuralink has said some decoders now hold up for three or more weeks without retraining, but the real test is whether that holds as more participants, with more varied brain anatomy and more varied daily routines, start relying on a Neuralink BCI for everyday tasks rather than supervised sessions.

Regulatory movement is worth watching too. Neuralink’s device currently has an FDA “breakthrough device” designation and operates under an investigational trial, not a general market clearance. Any future expansion to non-medical users would require a different regulatory path entirely, and the company hasn’t laid out a timeline for that beyond general statements of intent. In the meantime, competitors matter as a sanity check: if Synchron, Precision Neuroscience, or an academic program like BrainGate report comparable bits-per-second gains using a different approach, that’s a sign the recent jump reflects real progress in BCI software broadly, not just a one-off result specific to Neuralink’s hardware.

There’s also a quieter question worth tracking: privacy and data. A Neuralink BCI works by recording and interpreting brain activity continuously, and the pretraining approach behind this speed record depends on accumulating tens of thousands of hours of that data per person. Neither Neuralink nor its competitors have published detailed answers yet about how long that raw neural data is stored, who can access it, or what happens to it if a device is removed or a company changes hands — questions likely to get more attention as brain implants move from a handful of trial participants toward any kind of wider use.

Frequently Asked Questions

The 11.32-bits-per-second record will likely be broken again soon — that’s how this field has moved so far. What’s more significant is the trend underneath it: brain-computer interfaces are getting faster not just because of better hardware, but because of better software trained on more real-world data, the same pattern driving progress across AI more broadly. A Neuralink BCI that needed an hour of calibration a year ago now needs ten minutes, which is the kind of unglamorous, compounding improvement that tends to matter more over time than any single headline number. For the small number of people living with paralysis who rely on these implants day to day, that trend is the one actually worth watching.