01 / DECISIONFree public beta
Decision Model
Structured decisions from text and images: Choice, Multi-choice, Noul and Score. Select from up to 1024 candidates, with up to 10 multi-choice selections.
CONTEXT
Try model ⟶DECIDE
Surd AI is building the next generation of human–AI interaction systems. Powered by our Simplex model family, voice assistants, robots and digital humans bring natural, real-time communication and collaboration into everyday life.
Perception
Interaction
Memory
Spatial intelligence
Control
SURD
INTELLIGENCE
INTELLIGENCE
OUR VISION
Understanding the world.
Connecting people and AI.
We bring perception, conversation, memory and action into a unified framework, so AI can understand context, remember shared experiences and act alongside us. From voice assistants to robots and digital humans, we’re making natural, real-time interaction part of everyday life.
01
Understand
See the world in context.
02
Connect
Build on shared experiences.
03
Collaborate
Turn understanding into action.
SURD PRODUCT ECOSYSTEM
Simplex model family
Explore Simplex decision and voice models. The decision API is available to try now.
02 / VOICEComing soon
Voice Model
A model family for expressive, streaming speech. Model details, benchmarks and access information will be shared progressively.
Details and access coming soonTECHNOLOGY / NEWS
Technology Updates & News
Research · 2026.10.05
CMDB-1500: Multimodal Decision Model Benchmark
Compare model accuracy, probability calibration, and I/C indices on 1,200 text questions and 300 image questions.
↗Research · 2026.09.28Simplex SA0: A Full-Duplex Interaction Model for Voice, Action and Memory
Bringing action, speech, expression, memory and tool use into one continuous interaction loop.
↗Research · 2026.09.28Simplex P0: Understanding Human Intent Beyond the Wake Word
Combining gaze, speaking activity, expression and scene context to track events, social state and the right moment to engage—toward interaction without a wake word.
↗Research · 2026.09.25DeFM-WMP Research Preview: From Terrain Perception to Quadruped Locomotion
DeFM-WMP combines depth vision and a world model with locomotion control, adapting one quadruped policy across varied simulated terrain.
↗Engineering · 2026.09.24Introducing Simplex CD: our first general-purpose multimodal fast decision model
Make choices, select multiple options, estimate yes/no probabilities, and assign scores from text and images. Explore demos, CMDB-1500 results, and the open-source Omni model.
↗Research · 2026.09.17Simplex VM0: First-person experience for understanding and decisions
Connect visual, linguistic and spatial cues across time into retrievable, connected memories—a memory foundation for spatial intelligence, ongoing understanding and action.
↗Research · 2026.09.15Simplex MicroCF: #1 in interruption recall on TurnBench
Simplex MicroCF ranks first in interruption recall and third in end-of-turn recall on the TurnBench test set. Explore the results and its hybrid VAD + VAP architecture.
↗FOR DEVELOPERS & PARTNERS