What is the best AI model for real estate right now?
With the pace at which new AI models are being released, this has become a surprisingly difficult question to answer.
One week a model is best-in-class. The next week, a new release leapfrogs it. Pricing changes. Benchmark results update. Context windows expand. Open-weight models get better. And for those of us in commercial real estate, the right answer often depends on the task.
The best AI model for building a real estate financial model may not be the best model for reviewing leases. The best model for document review may not be the best model for a fast client-facing chat interface. And the best model for sensitive internal workflows may be one you can run privately.
So rather than try to answer the question once and have that answer go stale, we built a daily-updated LLM Leaderboard for real estate professionals.
The A.CRE LLM Leaderboard
Built by the AI.Edge by A.CRE team, the leaderboard below updates daily and is meant to help CRE professionals compare leading AI models using a practical set of metrics: intelligence, value, speed, cost, and key capabilities such as vision, reasoning, tools, long context, and open weight.
At the top of the leaderboard, you’ll see the editor’s picks for the most capable, best value, and fastest models. Below that, the full directory allows you to sort and filter based on what matters most for your use case.
How to Read the Leaderboard
The leaderboard is organized around four core metrics and five key capabilities.
Intelligence
Intelligence is a composite quality score that captures how “smart” a model is across a battery of standardized benchmarks, including reasoning, math, coding, and knowledge.
Higher numbers mean the model gets harder questions right more often. It is the closest single number to raw capability, which is why it is the default sort and why the hero card labeled “Most Capable” selects the top model on this metric.
Value
Value is a composite that rewards models offering the best mix of smarts and price within today’s leading tier.
Each day, we take the top 30 models by Intelligence Index that also have valid blended pricing, then normalize both metrics to a 0–100 scale within that cohort. Intelligence is scored linearly; cost (blended price per million tokens, weighted 3-to-1 input-to-output) is inverted, so the cheapest model in the cohort scores 100 and the most expensive scores 0.
Before scoring, we apply a quality floor: any model below the 40th percentile on intelligence within the cohort is excluded. A model that’s free but unreliable isn’t a good value for high-stakes work.
The remaining models are ranked by a weighted blend that favors being right over being cheap:
Value = (1.5 × intelligence + 1 × cost) / 2.5
Speed is deliberately left out, since for analytical workloads the difference between 50 and 200 tokens per second rarely matters.
These weights are an editorial choice tuned for our commercial real estate audience, not a universal optimum. This is the metric that powers the “Best Value” hero card.
Speed
Speed measures how quickly the model produces text, measured in output tokens per second.
Higher is snappier. This matters most for streaming chat interfaces, voice apps, and workflows where a user is watching the response generate in real time. Speed does not tell you anything about quality, but it matters a lot for user experience. This is the metric used for the “Fastest” hero card.
Cost
Cost is the blended price per million tokens, using the same 3-to-1 input-to-output weighting used in the Value calculation.
This gives you a single number to compare expected cost across models, rather than forcing you to separately evaluate input and output prices. Lower is cheaper, so this metric is sorted ascending.
Model Capabilities Explained
Vision
Vision means the model can accept images as input and reason about them.
In practical terms, this means the model can read text in screenshots, describe photos, parse charts, analyze UI mockups, and work with visual information. For CRE, this can be useful for reviewing screenshots, extracting information from marketing materials, analyzing charts, or working with documents that include visual elements.
Reasoning
Reasoning means the model has an explicit thinking mode where it performs extended internal deliberation before answering.
These models generally trade latency and cost for better performance on harder math, logic, planning, and multi-step problems. For simple Q&A, reasoning may be overkill. For complex problem-solving, financial modeling logic, or multi-step analysis, it can be transformative.
Tools
Tools means the model supports function or tool calling.
In other words, the model can request that an application run a function, such as searching the web, querying a database, sending an email, or running a calculation, and then incorporate the result into its response.
This is the foundation of agents, copilots, and AI applications where the model needs to take action rather than just produce text.
Long Context
Long context refers to the maximum number of tokens the model can hold in working memory at one time.
This includes the prompt, any documents attached, conversation history, and the model’s response. A 200K context window can fit a long novel. A 1M+ context window can fit a small codebase or hundreds of pages of documents.
For CRE, long context is especially useful for offering memorandums, leases, loan documents, market studies, and other long-form materials. Bigger context windows allow you to include more reference material without chunking, although quality can still degrade near the end of very large contexts.
Open Weight
Open weight means the model’s weights have been publicly released.
This allows you to download the model and run it on your own hardware, or through a hosting provider of your choice, rather than being locked into a single vendor’s API.
Open-weight models can offer advantages around privacy, cost control, fine-tuning, and vendor flexibility. Closed models, such as GPT, Claude, and Gemini, are accessed through the provider’s API or interface.
Which AI Model Should CRE Professionals Use?
There is no single best model for every real estate workflow.
For underwriting, financial modeling, and complex analysis, prioritize intelligence and reasoning. For high-volume document review, value and long context may matter more. For client-facing tools or chat interfaces, speed becomes more important. For sensitive internal workflows, open-weight models may be worth evaluating.
The goal of this leaderboard is not to crown one permanent winner. The goal is to help real estate professionals choose the right model for the task at hand.
Frequently Asked Questions about What is the Best AI Model for Real Estate Right Now?
Why is choosing the best AI model hard?
It is difficult because “new AI models are being released” quickly, and the best option can change as “pricing changes,” “benchmark results update,” “context windows expand,” and “open-weight models get better.” For commercial real estate, “the right answer often depends on the task.”
Is there one best model?
No. The post states, “There is no single best model for every real estate workflow.” It explains that “the best AI model for building a real estate financial model may not be the best model for reviewing leases,” and the best model for document review may differ from the best model for a “fast client-facing chat interface.”
What is the LLM Leaderboard?
The A.CRE LLM Leaderboard is a “daily-updated LLM Leaderboard for real estate professionals.” It is designed to help CRE professionals compare leading AI models using “intelligence, value, speed, cost, and key capabilities such as vision, reasoning, tools, long context, and open weight.”
How is intelligence measured?
Intelligence is a “composite quality score” that measures how “smart” a model is across standardized benchmarks, including “reasoning, math, coding, and knowledge.” Higher scores mean “the model gets harder questions right more often.”
What does value mean?
Value measures “intelligence per dollar.” Specifically, the post says it takes the Intelligence score and divides it by “the model’s blended price per million tokens,” using a “3-to-1 ratio of input to output.” This rewards models that “punch above their price point.”
When does speed matter?
Speed measures “how quickly the model produces text,” using output tokens per second. The post notes that speed matters most for “streaming chat interfaces, voice apps, and workflows where a user is watching the response generate in real time.”
What is long context?
Long context is “the maximum number of tokens the model can hold in working memory at one time,” including the prompt, attached documents, conversation history, and response. For CRE, it is useful for “offering memorandums, leases, loan documents, market studies, and other long-form materials.”
What are open-weight models?
Open weight means “the model’s weights have been publicly released.” This allows users to “download the model and run it on your own hardware, or through a hosting provider of your choice,” rather than relying on a single vendor’s API. The post says open-weight models can offer advantages around “privacy, cost control, fine-tuning, and vendor flexibility.”
How should CRE pros choose?
The post recommends matching the model to the task. For “underwriting, financial modeling, and complex analysis,” prioritize intelligence and reasoning. For “high-volume document review,” value and long context may matter more. For “client-facing tools or chat interfaces,” speed becomes more important. For “sensitive internal workflows,” open-weight models may be worth evaluating.
