Neuralink's New Data: ALS Patient Speaks via Brain Control at 75 Words Per Minute in First Week
On October 29, Neuralink, the brain-computer interface company backed by Elon Musk, released early language restoration data for patients with amyotrophic lateral sclerosis (ALS). The results show that patients implanted with the N1 device can now achieve voice-to-text and direct speech output at speeds close to normal conversation. One participant reached a typing speed of 75 words per minute in the first week after implantation, with a word error rate as low as 5.8% and a vocabulary coverage of 125,000 words. The data was presented by Dr. Nader Pouratian at the 2026 Northeast ALS Consortium (NEALS) annual meeting.
The results come from the VOICE clinical trial, registered under NCT07224256. This is an early feasibility study designed to evaluate the initial clinical safety and efficacy of the N1 implant and the R1 surgical robot. N1 is a skull-mounted, wireless, rechargeable brain-computer interface that uses the R1 robot to precisely insert ultra-fine electrode threads into the brain's speech motor cortex. When a person attempts to speak, these regions generate neural signals that normally travel to the muscles of the mouth, tongue, and vocal cords. Diseases such as ALS, stroke, spinal cord injury, cerebral palsy, and multiple sclerosis can block this pathway, leaving patients unable to produce intelligible speech. N1 records and decodes these "intent-to-speak" signals, turning them into text or synthesized voice.
The study mainly enrolls adults aged 22 or older who have severe speech impairment and limited upper-limb function, and who have a reliable caregiver. The current lead site is UT Southwestern Medical Center. Neuralink said the first N1 participant received the implant in January 2024, and since then multiple patients with ALS and spinal cord injury have been implanted. The company has already received FDA Breakthrough Device designation for speech restoration, with a long-term goal of reaching natural conversation speed — about 140 words per minute.
In the Chinese-language domain, speech decoding is harder because Mandarin carries tonal information and has a high density of homophones. Still, Chinese teams have made notable progress. In 2025, NeuroXess, together with Huashan Hospital affiliated with Fudan University and the Shanghai Institute of Microsystem and Information Technology of the Chinese Academy of Sciences, published real-time Mandarin decoding results in Science Advances. Using a 256-channel flexible implant system, after nine days of training, a participant achieved 71.2% accuracy in pure neural decoding across 394 common Mandarin syllables, with 65-millisecond single-syllable latency and a real-time sentence decoding speed of 49.6 characters per minute — enough for everyday communication.
Beijing Xinzhida's "Beinao-1" takes a semi-invasive approach. In 2025, an ALS patient with aphasia received the implant at Xuanwu Hospital. After only three hours of training, the patient could already output short phrases such as "I want to eat" and "Please call the doctor for me," with later accuracy improving to 52%. Single-character decoding latency is under 100 milliseconds, and the target speaking speed is above 80 characters per minute. Beinao-1 has now entered registered clinical trials, covering three patient groups: spinal cord injury, stroke, and ALS-related speech impairment.
The difference between Chinese and English means the numbers cannot be compared directly. English is measured by words per minute, while Chinese is measured by characters per minute, and Mandarin decoding is inherently more difficult. Neuralink's 75 words per minute represents high real-time performance in an English setting, while NeuroXess's 49.6 characters per minute is the peer-reviewed benchmark for pure neural decoding in Mandarin. With large language model assistance, Chinese systems can reach higher engineering speeds, but clinical endpoints still rely on the former.
Despite the excitement around Neuralink and other companies, these devices remain research-stage technologies. Neuralink emphasizes that its system has not yet received approval for commercial use from the FDA or any other regulatory body. Patients or family members interested in learning more can join the Patient Registry through the company's website for trial details and participation pathways. Speech-restoring brain-computer interfaces are moving from the lab toward the clinic, but further validation and iteration are still needed.
