Tether EVO earns dual top ranking in brain AI benchmark

Tether said its “EVO Engineers” team placed fourth and a separate “Tether Evo” entry placed fifth on the Brain-to-Text ’25 leaderboard, a contest focused on decoding speech intent from neural activity. The company described the result as validation for a “local-first” approach designed to run without continuous cloud connectivity, emphasizing low latency, privacy, and robustness to noisy inputs.

The Brain-to-Text ’25 challenge requires participants to translate 256 channels of raw neural activity into readable text without precise time-alignment data, using Electrocorticography (ECoG) recordings. Tether said its methods aim to represent high-dimensional neural signals efficiently and remain usable where infrastructure constraints make remote inference impractical.

Paolo Ardoino, Tether’s CEO, said the company views brain-computer interfaces as part of an effort to keep “full control” in users’ hands rather than in centralized data centers. He framed the competition result as both an engineering milestone and a proof point for building AI systems that prioritize privacy and self-sovereignty.

The benchmark sits inside a growing research and engineering push to decode language from the brain, a line of work that is often positioned as a future communication pathway for people who cannot speak. Organizers of the Brain-to-Text benchmark describe the goal as improving algorithms that decode speech from brain activity, with accuracy improvements tied to practical use. 

Academic write-ups around the earlier Brain-to-Text Benchmark ’24 competition show how evaluation design is evolving to better reflect real deployment conditions, including dataset scale and the interaction between neural decoding models and language modeling. Tether’s Feb. 12 statement echoed that framing by highlighting that its submission performed under “real-world simulated constraints,” rather than an idealized, cloud-heavy setup.

Tether’s announcement also arrives as the company broadens its public messaging beyond stablecoins into technology initiatives branded under “EVO.” In a separate update one day earlier, on Feb. 11, 2026, Tether launched a public “USD₮ Tether Directory” intended to map where its USD₮ stablecoin is used across exchanges, payments and wallets, describing the tool as a transparency effort for users, developers, institutions and regulators.

Tether did not disclose technical details such as model architectures, compute budgets, or whether its top-five entries were trained and evaluated entirely on local hardware. The company’s statement focused on design priorities—local execution and robustness—and on the contest’s input format and constraints.

The leaderboard snapshot for Brain-to-Text ’25 shows the two Tether-related entries clustered near the top, with “Tether Evo Engineers” listed fourth and “Tether Evo” listed fifth by score. While the competition is one benchmark, results in this space are closely watched because small improvements in decoding accuracy and stability can compound into meaningful gains for assistive communication systems.

For now, Tether is using the result to position EVO as a unit working at the intersection of biology and machine intelligence, describing its focus areas as brain-computer interfaces and neuroprosthetics. The next signals the market will watch are whether Tether publishes more detail on how its models achieved top-five performance, and whether it translates benchmark credibility into partnerships, prototypes, or published research that can be independently evaluated.