Synthetic Sciences Releases OpenScience — Open AI Workbench for ML, Biology, Physics and Chemistry

Synthetic Sciences released OpenScience on July 5, 2026 — an open-source AI workbench under the Apache 2.0 license for automating research in machine learning, biology, physics, and chemistry. The platform works with any frontier or open-weight model through the researcher's API keys and is deployed on the organization's own infrastructure — without transmitting data to external services.

What's included in the standard release

The workbench covers the full research cycle — from searching scientific databases to executing computational tasks — without requiring switching between incompatible tools from different providers.

Key components of the release:

• License — Apache 2.0, source code fully open
• Coverage: machine learning, biology, physics, chemistry
• Over 250 editable skills in the standard release
• Built-in access to requested scientific databases
• Compatibility with any frontier or open-weight model through user API keys
• Deployment on the organization's own infrastructure

Each skill is an editable unit of logic. A researcher can rewrite a prompt, change the sequence of steps, or add specific checks without reworking the entire platform. This fundamentally distinguishes OpenScience from SaaS tools, where users are constrained by vendor-approved options: here, the laboratory adapts ready-made solutions to its task, rather than fitting the task to tool limitations.

Why is independence from AI models important?

The core problem stems from the lock-in of most AI tools for science to a single provider. Switching models means switching the entire work environment: a different platform, new prompts, team retraining. This hinders experimentation with new models and creates long-term vendor dependency.

OpenScience breaks this dependency through a unified interface over any engine. A researcher chooses a model for the task: a powerful frontier model for complex multi-step reasoning, or a lightweight open model for routine data processing and cost optimization. Academic groups often work with multiple providers, switching depending on grants and institutional licenses. Model neutrality protects against vendor lock-in and enables seamless transitions to new models as they emerge.

A separate concern is data confidentiality. Biological experiments, chemical compounds, and unpublished genomic sequences often contain information that organizations cannot transmit to external cloud services. OpenScience enables the entire research cycle to take place within the laboratory's perimeter — from problem formulation to results. This removes regulatory barriers that blocked many institutions from adopting external AI services.

Scientific databases inside the pipeline

Built-in access to scientific databases is one of the platform's key announced features. AI skills can directly access structured scientific data within the research pipeline, bypassing manual export and dataset upload. This reduces routine preparatory work in literature search, training sample selection, and result verification.

Coverage of four disciplines — ML, biology, physics, and chemistry — reflects orientation toward the AI4Science direction, where AI is applied to accelerate scientific discoveries: from protein structure prediction to quantum chemistry tasks. A universal open workbench aims to become the infrastructure layer linking experimental laboratories with current AI capabilities.

What this means

OpenScience is a bid for an open standard in AI tools for natural sciences. A combination of the Apache 2.0 license, model neutrality, 250+ editable skills, and local deployment targets academic institutions and corporate R&D laboratories for which vendor lock-in and data confidentiality requirements are real blockers. How widely the scientific community adopts the platform will be evident from repository contribution activity in the coming months.