Cadence unveils Maestro AI models for chip design, offering early access in Q4 2026

Cadence Introduces Maestro AI

Cadence, an electronic design automation (EDA) company, announced Maestro, a new portfolio of domain-specialized AI models. These models are designed to work with Cadence's chip design agents.

Initial results indicate up to a 30% improvement in the quality of results and up to 5x token efficiency when compared to general-purpose AI models for the same tasks.

Availability and Deployment

The first Maestro models will be available to select early access customers in Q4 2026. Customers will have the flexibility to run these models on Cadence’s cloud, public clouds, or their own on-premises environments.

Users can also post-train the models using their proprietary data, designs, and workflows.

Distinction from Competitors

Cadence's approach with Maestro differs from Synopsys' GPT-Synopsys, which operates on OpenAI-hosted infrastructure without using customer data for training. This highlights a divergence in how the two EDA giants handle customer design data.

Paul Cunningham, SVP and GM at Cadence, noted that agent costs have increased, making it logical to integrate more of the agents' iterative processes into the models' weights.

Maestro's Operational Mechanism

Maestro functions through Cadence’s AI Super Agents, which interpret design intent, break down tasks, and orchestrate the Maestro portfolio with other frontier and open-weight models. These agents evaluate trade-offs to guide design decisions using domain expertise.

The system also employs reinforcement learning, where Cadence’s deterministic solvers assess outputs against engineering objectives, constraints, and ground truth, with results feeding back into the models for continuous improvement. The models include Nvidia’s Nemotron, and customers can perform the customer-data training step using Cadence’s fine-tuning harness.