Reasoner

Human-level reasoning is the holy grail of AI.

“Can I trust it?
We don’t expect hallucinations to go away.”

Bret Greenstein
Principal, Data and Analytics

“Ensuring a level of transparency on the algorithmic decision-making process is critical... and the black box problem is just one of many risks...”

Bhaskar Chakravorti
Dean of Global Business at Tufts University

“We have to get to a point where the answer that you get you largely trust and I think that we're several years away from being able to do that.”

JENSEN HUANG
CEO

Our mission is to bring human-level reasoning to complex enterprise applications and workflows that demand the highest precision and accuracy.

The first accurate neurosymbolic engine

Knowledge consistency that scales

The data is clear. As complexity rises, accuracy declines—except with Reasoner.

Actual reasoning,
clear results.

When it comes to reasoning, Reasoner delivers clarity where others fall short.

Here’s an example: The surgeon riddle

The surgeon, who is the boy's father, says, "I can't operate on this boy, he's my son!" Who is the surgeon to the boy?

Thinking

Thought for 23 seconds

The surgeon is the boy's grandfather.

Here's the reasoning:

The surgeon is described as the boy's father.

When the surgeon says, "I can't operate on this boy, he's my son!", it implies that the boy is his son.
Therefore, the boy's father (the surgeon) refers to the boy as "my son."

This means that the surgeon is both the boy's father and refers to the boy's father as "my son," indicating that the surgeon is actually the boy's grandfather.

Answer: His grandfather—the surgeon is the boy’s grandfather.

Reasoning

Reasoned

Data Analysis

steps action

ANALYZE STATEMENT

I’ll analyze the surgeon’s reason for not performing the surgery. This will reveal the surgeon-boy relationship.

Reasoning: This analysis provides the first piece of evidence. It forms the foundation for my conclusion and sets up the verification step.

Plan Execution: Examined: "I can't operate on this boy, he's my son!"
Breakdown:
1. Can’t operate
2. Reason: boy is son
Inference: Paternal relationship

VERIFY RELATIONSHIP

I’ll compare findings from step 1 with the statement "The surgeon is the boy's father." If these conflict, I’ll re-analyze the surgeon's statement.

Reasoning: Cross-referencing increases the reliability of my conclusion. It ensures support from multiple evidence pieces.

Plan Execution: Verified with explicit statement: "The surgeon, who is the boy's father."
Relevant Information Found: Direct confirmation of initial inference.

DRAW CONCLUSION

I'll clearly state the surgeon-boy relationship based on the given information.

Reasoning: This step brings together all analyzed information. Basing my conclusion on multiple, consistent pieces of evidence ensures high reliability.

Plan Execution: Synthesized all information.
Validation: Implied and explicit info align perfectly.
Conclusion: Surgeon is boy’s father.

Reasoning

Reasoned

Data Analysis

Answer: The surgeon is the boy's father.

Here’s the explanation:

Based on the provided information, the relationship between the surgeon and the boy is clearly and unequivocally established.

The surgeon's statement, "I can't operate on this boy, he's my son!" combined with the explicit statement that "The surgeon is the boy's father," leads to the definitive conclusion that the surgeon is the boy's father.

The puzzle plays on the assumption that a surgeon is male, but the statement directly confirms the father-son relationship.

Benchmark-beating reasoning

A data-driven analysis of Reasoner’s engine’s performance using Google’s FRAMES benchmark shows Reasoner excels at factuality, retrieval accuracy, and complex reasoning.

77.3%

ACCURACY

92.8%

accuracy achieved

847

tasks correct

913

Total Reasoning Tasks

1,647

Total Sources Used

Outperforming in every category

Reasoner's engine delivers unmatched results across all types of reasoning.

Assessment Type	TOTAL ASSESSMENTS	REASONER % Raw Score	OPENAI o1 % Raw Score
Numerical Reasoning	121	93.5	71.1
Tabular Reasoning	151	87.5	78.1
Multiple Constraints	400	93.5	79.3
POST PROCESSING	71	94.4	73.2
TEMPORAL Reasoning	170	94.7	78.2
Total	913	92.8	77.3