Sentient, a San Francisco-based AI development lab valued at $1.2 billion, announced Tuesday afternoon the public release of Open Deep Search (ODS), an open-source AI search framework.
Over key benchmarks, Sentient claims it outperforms major closed-source competitors like Perplexity and OpenAI’s GPT-4o Search Preview.
Backed by Peter Thiel’s Founder’s Fund, Sentient claims its work represents America’s answer to China’s DeepSeek.
When it broke out, DeepSeek’s open-source model questioned industry beliefs and caused a $1 trillion tech slump.
Sentient’s models challenge the “dominance of closed AI systems” as the U.S. approaches its own “DeepSeek moment,” the company said in a statement shared with Decrypt.
Operating as a nonprofit, Sentient argues AI development “should belong to the community, not controlled by closed-source corporations.”
It’s a numbers game
According to Sentient testing, ODS scored 75.3% accuracy on the FRAMES benchmark, significantly outperforming GPT-4o Search Preview (50.5%) and Perplexity Sonar Reasoning Pro (44.4%).
ODS uses Sentient’s Open Search Tool, powered by an agentic framework, to rephrase queries and extract context from results.
Asked about their testing methodology, Sentient co-founder Himanshu Tyagi told Decrypt that the benchmarking framework required models “to orchestrate knowledge from multiple sources.”
To make ODS more efficient, “ground truth” sources such as Wikipedia were excluded. This ensured that models “relied on their retrieval systems” instead of being artificially assisted, creating “a more realistic and rigorous evaluation,” Tyagi claimed.
ODS reached these numbers using an “agentic approach that writes self-correcting code,” Tyagi said. That approach helped build out and determine which questions were needed to get a final answer.
When the framework misses a critical piece of info, it calls the search tool again, but with a more specific query to retrieve more precise information.
If it hits a speed bump, the model does “enhanced query rephrasing, multi-pass retrieval, and intelligent chunking and reranking of web pages,” Tyagi explained.
ODS and its evaluations can now be reviewed on their public GitHub repository.
America’s ‘DeepSeek moment’
When DeepSeek came to global attention in January, Decrypt asked NYU Shanghai professor Bogna Konior why this was a key moment.
“We now routinely let AI draft our thoughts—a development as remarkable as the invention of language itself,” Konior told Decrypt. “It’s as if humanity is recreating that pivotal moment of language invention within computers.”
The parallels between DeepSeek and Sentient show these philosophical shifts for AI.
“Once open-source technology is released into the world, it cannot be contained,” Konior said.
Sentient believes that the moment has arrived for America.
Edited by Sebastian Sinclair
Generally Intelligent Newsletter
A weekly AI journey narrated by Gen, a generative AI model.
