Proprietary LLM: Luna Crescent
Luminance’s proprietary LLM is purpose-built for specialist legal tasks, improving the speed and economics of high-volume contract extraction, classification and analysis.
Proprietary LLM: Luna Crescent
Luminance’s proprietary LLM is purpose-built for specialist legal tasks, improving the speed and economics of high-volume contract extraction, classification and analysis.
Proprietary LLM: Luna Crescent
Luminance’s proprietary LLM is purpose-built for specialist legal tasks, improving the speed and economics of high-volume contract extraction, classification and analysis.
3.4M
Legal data points learned to date
189K+
data point benchmark test
4x
faster than general-purpose frontier models on Contract Intelligence tasks
5%
higher accuracy in Blind Legal Expert Tests
3.4M
Legal data points learned to date
189K+
data point benchmark test
4x
faster than general-purpose frontier models on Contract Intelligence tasks
5%
higher accuracy in Blind Legal Expert Tests
3.4M
Legal data points learned to date
189K+
data point benchmark test
4x
faster than general-purpose frontier models on Contract Intelligence tasks
INTRODUCING LUNA CRESCENT
Legal intelligence built for contracts
Luna Crescent is Luminance’s family of proprietary legal intelligence models, trained to understand legal language, contractual structure, and the relationships that determine commercial meaning.
Luna Crescent combines proprietary legal data, specialist training, a rigorous benchmark and infrastructure owned by Luminance. Refined over years of iteration, it delivers a depth of legal understanding, control and performance that goes far beyond the underlying foundation model.
By combining Luna Crescent with the latest frontier models, Luminance delivers faster contract understanding, more reliable legal intelligence and a platform that can continually improve as AI advances.
INTRODUCING LUNA CRESCENT
Legal intelligence built for contracts
Luna Crescent is Luminance’s family of proprietary legal intelligence models, trained to understand legal language, contractual structure, and the relationships that determine commercial meaning.
Luna Crescent combines proprietary legal data, specialist training, a rigorous benchmark and infrastructure owned by Luminance. Refined over years of iteration, it delivers a depth of legal understanding, control and performance that goes far beyond the underlying foundation model.
By combining Luna Crescent with the latest frontier models, Luminance delivers faster contract understanding, more reliable legal intelligence and a platform that can continually improve as AI advances.
INTRODUCING LUNA CRESCENT
Legal intelligence built for contracts
Luna Crescent is Luminance’s family of proprietary legal intelligence models, trained to understand legal language, contractual structure, and the relationships that determine commercial meaning.
Luna Crescent combines proprietary legal data, specialist training, a rigorous benchmark and infrastructure owned by Luminance. Refined over years of iteration, it delivers a depth of legal understanding, control and performance that goes far beyond the underlying foundation model.
By combining Luna Crescent with the latest frontier models, Luminance delivers faster contract understanding, more reliable legal intelligence and a platform that can continually improve as AI advances.
Expertise built in
Legal expertise embedded in the model
Luna Crescent begins with a capable general-purpose base model and is then trained on Luminance’s proprietary contract data. This specialist training teaches Luna Crescent to understand legal language, contractual structure and the relationships between clauses, obligations and commercial terms.
Breadth of Training Data: 3.4M+ data points have been extracted to create the breadth of Luna Crescent’s training data reflecting the reality of enterprise contracts.
Built for Diverse Enterprise Contracts: Training data encompasses different industries, difficult contract formats, obscure drafting styles and varied legal language.
Specialization that is Proprietary: The underlying model may be available to others, but the legal training, proprietary data and resulting specialist model belong to Luminance.
Expertise built in
Legal expertise embedded in the model
Luna Crescent begins with a capable general-purpose base model and is then trained on Luminance’s proprietary contract data. This specialist training teaches Luna Crescent to understand legal language, contractual structure and the relationships between clauses, obligations and commercial terms.
Breadth of Training Data: 3.4M+ data points have been extracted to create the breadth of Luna Crescent’s training data reflecting the reality of enterprise contracts.
Built for Diverse Enterprise Contracts: Training data encompasses different industries, difficult contract formats, obscure drafting styles and varied legal language.
Specialization that is Proprietary: The underlying model may be available to others, but the legal training, proprietary data and resulting specialist model belong to Luminance.
Expertise built in
Legal expertise embedded in the model
Luna Crescent begins with a capable general-purpose base model and is then trained on Luminance’s proprietary contract data. This specialist training teaches Luna Crescent to understand legal language, contractual structure and the relationships between clauses, obligations and commercial terms.
Breadth of Training Data: 3.4M+ data points have been extracted to create the breadth of Luna Crescent’s training data reflecting the reality of enterprise contracts.
Built for Diverse Enterprise Contracts: Training data encompasses different industries, difficult contract formats, obscure drafting styles and varied legal language.
Specialization that is Proprietary: The underlying model may be available to others, but the legal training, proprietary data and resulting specialist model belong to Luminance.
Reliable by design
Understand contracts within context
For legal work, trust matters more than fluency. Luna Crescent is designed to remain faithful to the underlying contract, distinguish between information that is present and information that is absent, and interpret legal language as a connected system of rights, obligations and risks.
Faithful to the source: Luna Crescent is trained to extract what the contract actually says rather than fill gaps with a plausible sounding answer.
Flags what is absent: When the requested information is not present in the contract, Luna Crescent is trained to identify that absence rather than invent an answer
Reads legal relationships: Luna Crescent considers how clauses, definitions, exceptions and obligations relate to one another rather than treating each sentence as isolated prose.
Reliable by design
Understand contracts within context
For legal work, trust matters more than fluency. Luna Crescent is designed to remain faithful to the underlying contract, distinguish between information that is present and information that is absent, and interpret legal language as a connected system of rights, obligations and risks.
Faithful to the source: Luna Crescent is trained to extract what the contract actually says rather than fill gaps with a plausible sounding answer.
Flags what is absent: When the requested information is not present in the contract, Luna Crescent is trained to identify that absence rather than invent an answer
Reads legal relationships: Luna Crescent considers how clauses, definitions, exceptions and obligations relate to one another rather than treating each sentence as isolated prose.
Reliable by design
Understand contracts within context
For legal work, trust matters more than fluency. Luna Crescent is designed to remain faithful to the underlying contract, distinguish between information that is present and information that is absent, and interpret legal language as a connected system of rights, obligations and risks.
Faithful to the source: Luna Crescent is trained to extract what the contract actually says rather than fill gaps with a plausible sounding answer.
Flags what is absent: When the requested information is not present in the contract, Luna Crescent is trained to identify that absence rather than invent an answer
Reads legal relationships: Luna Crescent considers how clauses, definitions, exceptions and obligations relate to one another rather than treating each sentence as isolated prose.
Proven before trusted
Tested before it reaches production
Rigorously benchmarked: ContractIQ Bench contains 189,000 manually annotated and reviewed data points covering real provisions such as liability caps, termination rights and confidentiality obligations. Luna Crescent is evaluated on concepts it has not encountered during training, providing a genuine test of whether it can apply its legal understanding to new material. Legal analysts and specialists review Luna Crescent’s output in blind and A/B testing, including previously unseen concepts from complex agreements.
Compared against leading models: Luna Crescent’s performance is assessed against leading general-purpose models rather than solely against its own previous versions.
Gated before release: No new version is promoted into production until it clears the benchmark and passes expert review.
Proven before trusted
Tested before it reaches production
Rigorously benchmarked: ContractIQ Bench contains 189,000 manually annotated and reviewed data points covering real provisions such as liability caps, termination rights and confidentiality obligations. Luna Crescent is evaluated on concepts it has not encountered during training, providing a genuine test of whether it can apply its legal understanding to new material. Legal analysts and specialists review Luna Crescent’s output in blind and A/B testing, including previously unseen concepts from complex agreements.
Compared against leading models: Luna Crescent’s performance is assessed against leading general-purpose models rather than solely against its own previous versions.
Gated before release: No new version is promoted into production until it clears the benchmark and passes expert review.
Proven before trusted
Tested before it reaches production
Rigorously benchmarked: ContractIQ Bench contains 189,000 manually annotated and reviewed data points covering real provisions such as liability caps, termination rights and confidentiality obligations. Luna Crescent is evaluated on concepts it has not encountered during training, providing a genuine test of whether it can apply its legal understanding to new material. Legal analysts and specialists review Luna Crescent’s output in blind and A/B testing, including previously unseen concepts from complex agreements.
Compared against leading models: Luna Crescent’s performance is assessed against leading general-purpose models rather than solely against its own previous versions.
Gated before release: No new version is promoted into production until it clears the benchmark and passes expert review.
Built to improve
Every lesson makes the next version better
Luna Crescent runs on infrastructure controlled by Luminance. That gives Luminance control over how the model is tested, deployed and improved, rather than depending entirely on the priorities, pricing, or availability of a third-party model provider.
The training corpus expands; the model is retrained, and each new version must outperform the one already in production before it is released.
Luna Crescent's Improvement Cycle:
Grow: Expand and refine the training corpus with more varied and challenging contract data.
Retrain: Train a new version using the enlarged and improved dataset.
Benchmark: Evaluate it against held-out concepts and the strongest currently available models.
Validate: Have legal experts review the output on real, unseen work.
Promote: Release the new version only when it genuinely improves the version already in production.
Built to improve
Every lesson makes the next version better
Luna Crescent runs on infrastructure controlled by Luminance. That gives Luminance control over how the model is tested, deployed and improved, rather than depending entirely on the priorities, pricing, or availability of a third-party model provider.
The training corpus expands; the model is retrained, and each new version must outperform the one already in production before it is released.
Luna Crescent's Improvement Cycle:
Grow: Expand and refine the training corpus with more varied and challenging contract data.
Retrain: Train a new version using the enlarged and improved dataset.
Benchmark: Evaluate it against held-out concepts and the strongest currently available models.
Validate: Have legal experts review the output on real, unseen work.
Promote: Release the new version only when it genuinely improves the version already in production.
Built to improve
Every lesson makes the next version better
Luna Crescent runs on infrastructure controlled by Luminance. That gives Luminance control over how the model is tested, deployed and improved, rather than depending entirely on the priorities, pricing, or availability of a third-party model provider.
The training corpus expands; the model is retrained, and each new version must outperform the one already in production before it is released.
Luna Crescent's Improvement Cycle:
Grow: Expand and refine the training corpus with more varied and challenging contract data.
Retrain: Train a new version using the enlarged and improved dataset.
Benchmark: Evaluate it against held-out concepts and the strongest currently available models.
Validate: Have legal experts review the output on real, unseen work.
Promote: Release the new version only when it genuinely improves the version already in production.
the luminance approach
The right model for every task
Different tasks require different strengths. Luna Crescent isn’t intended to replace frontier models or force every workflow through a single proprietary model.
Luna Crescent:
Works alongside leading general-purpose models within Luminance’s multi-model architecture, which automatically directs each task to the model best suited to perform it.
Handles high-volume work where specialist legal understanding, speed and repeatability provide the greatest advantage. Frontier models are used where broader generative capabilities remain stronger.
the luminance approach
The right model for every task
Different tasks require different strengths. Luna Crescent isn’t intended to replace frontier models or force every workflow through a single proprietary model.
Luna Crescent:
Works alongside leading general-purpose models within Luminance’s multi-model architecture, which automatically directs each task to the model best suited to perform it.
Handles high-volume work where specialist legal understanding, speed and repeatability provide the greatest advantage. Frontier models are used where broader generative capabilities remain stronger.
the luminance approach
The right model for every task
Different tasks require different strengths. Luna Crescent isn’t intended to replace frontier models or force every workflow through a single proprietary model.
Luna Crescent:
Works alongside leading general-purpose models within Luminance’s multi-model architecture, which automatically directs each task to the model best suited to perform it.
Handles high-volume work where specialist legal understanding, speed and repeatability provide the greatest advantage. Frontier models are used where broader generative capabilities remain stronger.
Explore the architecture
From the right context to the right answer.
Document Search Engine
Pre-computed legal knowledge and Context Graphs connect contracts, amendments and related documents, enabling users and AI agents to retrieve the right evidence faster and analyse transactions in context.
Model-Opinionated
A multi-model architecture routes each task to the intelligence best suited to its accuracy, speed, reliability and cost requirements, without dependence on a single model or provider.
Differential Validation
Independent judges, grounded critics and task-specific checks scrutinise high-impact outputs against evidence, improving accuracy and making uncertainty visible when human review is required.
Explore the architecture
From the right context to the right answer.
Document Search Engine / Context Graphs
Pre-computed legal knowledge and Context Graphs connect contracts, amendments and related documents, enabling users and AI agents to retrieve the right evidence faster and analyse transactions in context.
Model-Opinionated
A multi-model architecture routes each task to the intelligence best suited to its accuracy, speed, reliability and cost requirements, without dependence on a single model or provider.
Differential Validation
Independent judges, grounded critics and task-specific checks scrutinise high-impact outputs against evidence, improving accuracy and making uncertainty visible when human review is required.
Explore the architecture
From the right context to the right answer.
Document Search Engine / Context Graphs
Pre-computed legal knowledge and Context Graphs connect contracts, amendments and related documents, enabling users and AI agents to retrieve the right evidence faster and analyse transactions in context.
Model-Opinionated
A multi-model architecture routes each task to the intelligence best suited to its accuracy, speed, reliability and cost requirements, without dependence on a single model or provider.
Differential Validation
Independent judges, grounded critics and task-specific checks scrutinise high-impact outputs against evidence, improving accuracy and making uncertainty visible when human review is required.
GET STARTED
Get more from your contracts
See how Luminance can help your teams negotiate smarter, surface what matters, and keep contracts moving across the enterprise.
GET STARTED
Get more from your contracts
See how Luminance can help your teams negotiate smarter, surface what matters, and keep contracts moving across the enterprise.
GET STARTED
Get more from your contracts
See how Luminance can help your teams negotiate smarter, surface what matters, and keep contracts moving across the enterprise.
© 2026 Luminance. All rights reserved.