Not Diamond

Not Diamond is a sophisticated AI infrastructure platform designed for developers and enterprises navigating the complex, multi-model future of artificial intelligence. It operates as a "meta-model" or an intelligent routing layer that sits on top of various large language models (LLMs). The core mission of Not Diamond is to enable teams to leverage the strengths of multiple AI models, rather than being locked into a single one. By doing so, it dramatically improves accuracy, reduces operational costs, and lowers latency for AI-powered applications.

The platform is built on the belief that the future of AI is not a single, monolithic model, but a networked ecosystem of thousands of specialized models. Not Diamond provides the critical infrastructure to manage this ecosystem efficiently. It analyzes incoming requests and, based on user-defined evaluation data and business logic, predictively determines which model (e.g., GPT-4o, Claude 3.5 Sonnet, etc.) is best suited for the task. This process outperforms any individual model in terms of quality while simultaneously optimizing for speed and cost.

How to use Not Diamond

Not Diamond is designed for seamless integration into existing development workflows. Developers can access its capabilities through a Python SDK, a TypeScript client, or a flexible REST API, making it compatible with virtually any tech stack.

The typical workflow is as follows:

1. Integration: Integrate the Not Diamond SDK or API into your application.
2. Training the Router: Provide your own evaluation data (examples of inputs and desired outputs or quality scores). Not Diamond uses this data to train a custom router that understands your specific domain, quality definitions, and business requirements.
3. Routing Requests: Instead of calling an LLM API directly, you send your request to the Not Diamond endpoint.
4. Receiving Recommendations: Not Diamond's router analyzes the input and, in under 60ms, returns a recommendation for the optimal model to use for that specific request.
5. Client-Side Execution: Your application then makes a direct, client-side call to the recommended LLM's API. Not Diamond is not a proxy; it only provides the routing intelligence, ensuring you maintain full control over your data and API calls.

Core Features of Not Diamond

• Intelligent Model Routing: Leverages your evaluation data to predictively select the most suitable LLM for each input, balancing accuracy, cost, and latency to outperform any single model.
• Automatic Prompt Adaptation: Automatically translates and optimizes a prompt written for one model (e.g., GPT-4o) to perform effectively on any other target model (e.g., Claude 3.5 Sonnet), saving significant manual prompt engineering time.
• Steerable Tradeoffs: Allows you to set quality thresholds, enabling the system to use faster and cheaper models for less critical tasks without compromising the quality of the output.
• Blazing-Fast Inference: The model selection process is incredibly fast, typically taking under 50-60ms, which means it adds negligible overhead and can even lead to net speedups by routing to faster models.
• Enterprise-Grade Security: Not Diamond is SOC-2 compliant and offers robust security features, including client-side request execution, a zero data retention policy, and support for VPC deployments for maximum security at scale.
• Developer-Friendly SDKs: Provides easy-to-use clients for Python and TypeScript, as well as a comprehensive REST API for integration into any environment.

Use Cases for Not Diamond

Not Diamond is ideal for teams scaling their AI operations across multiple applications and models.

• Customer Support Automation: A company can route simple, high-volume customer queries to a fast, inexpensive model, while complex, nuanced support requests are sent to a more powerful, state-of-the-art model, optimizing both cost and customer satisfaction.
• Content Generation Platforms: A marketing platform can use Not Diamond to select the best model for different content types—a creative model for blog posts, a concise model for social media updates, and a technical model for documentation.
• Data Analysis and Summarization: An application that analyzes various documents can route requests to summarize a legal contract to a model with strong reasoning capabilities, while routing a request to summarize a news article to a faster, more general-purpose model.

Advantages of Not Diamond

By adopting a multi-model strategy with Not Diamond, development teams gain significant competitive advantages:

• Cost Reduction: Drastically lower LLM operational costs by avoiding overuse of expensive models for tasks that can be handled by cheaper alternatives.
• Improved Accuracy & Quality: Achieve higher overall output quality by always using the best tool for the job, leveraging the unique strengths of different models.
• Accelerated Development: Radically shorten development cycles by abstracting away the complexity of model selection and prompt engineering.
• Reduced Latency: Achieve faster response times by intelligently routing to quicker models when appropriate.
• Future-Proofing: Easily adapt to the rapidly evolving AI landscape by seamlessly integrating new models without refactoring existing application logic.

Pricing and Plans

Not Diamond offers a tiered pricing structure to suit teams of all sizes:

• Discovery (Free): Includes up to 100,000 monthly API routing requests, the ability to train one custom router, and support for prompt adaptation. Ideal for individuals and small teams starting out.
• Possibility ($100/month + usage): Includes everything in the Discovery plan, plus uncapped API routing requests (at $0.001 per request after the first 100K), unlimited custom routers, and enhanced data privacy features. Designed for growing teams and applications.
• Necessity (Custom Pricing): A bespoke plan for large enterprises. It includes all features of the Possibility plan, plus VPC deployments, dedicated integration and training support, and advanced access and permissions management.