Contextual AI’s new AI model crushes GPT-4O in accuracy this is why it matters

Contextual AI reveal Targeted language model (GLM) Today that it claims that it provides the highest factual accuracy in industry by leading leading AI systems of Google” Anthropic And Openi On an important benchmark for truthfulness.

The startup, founded by the pioneers of Pick-up-advanced generation (RAG) Technology, reported that the GLM achieved an 88% actual score on the Facts benchmarkCompared to 84.6% for Google’s Gemini 2.0 Flash79.4% for that of Anthropic Claude 3.5 Sonnet and 78.8% for OpenAi’s GPT-4O.

Although large language models have transformed Enterprise software, factual inaccuracies – often called hallucinations – remain a crucial challenge for the acceptance of business activities. Contextual AI wants to solve this by creating a model that is specifically optimized for Enterprise RAG applications where accuracy is of the utmost importance.

“We knew that part of the solution would be a technique with the name RAG-OPEN-OTD generation,” said Douwe Kiela, CEO and co-founder of contextual AI, in an exclusive interview with Venturebeat. “And we knew that because Rag was originally my idea. What this company is about is really about doing RAG in the right way, to the next level of doing day. “

The company’s focus differs considerably from general models such as Chatgpt or ClamberThey are designed to handle everything, from creative writing to technical documentation. Instead, contextual AI focuses on Enterprise environments where factual precision outweighs creative flexibility.

“If you have a raging problem and you are in a company in a very regulated industry, you have no tolerance for hallucination,” Kiela explained. “The same general language model that is useful for the marketing department is not what you want in a business institution where you are much more sensitive to mistakes.”

How contextual ai ‘orientation’ makes the new gold standard for business models

The concept of “thoroughness” – Make sure that AI answers strictly stick to information that is explicitly provided in the context – has emerged as a critical requirement for AI systems from Enterprise. In regulated industries such as finance, health care and telecommunications, companies need AI that provides accurate information or explicitly recognizes when it does not know something.

Kiela offered an example of how this strict surface works: “If you give a recipe or formula to a standard language model, and somewhere in, you say:” But this is only true for most cases, “most language models will still give you the recipe that it is true. But our language model says: “It really only says that this applies to most cases.” It records this extra little nuance. “

The possibility to say “I don’t know” is crucial for Enterprise settings. “That is really a very powerful function, if you think about it in an enterprise setting,” Kiela added.

Contextual AIs RAG 2.0: A more integrated way to process company information

The Platform of Contextual AI is built on what it calls “RAG 2.0‘An approach that goes beyond the simple connection of ready-made components.

“A typical raging system uses a frozen ready-made model for embedding, a vector database for retrieval and a Black-box language model for generation, filled by prompt or an orchestra framework,” said a business statement. “This leads to a ‘Frankenstein’s monster’ from generative AI: the individual components work technically, but the whole is far from optimal.”

Instead, contextual AI jointly optimizes all the components of the system. “We have this component of the mixture of retrics, which is really a way to get intelligent pick up,” Kiela explained. “It looks at the question, and then it thinks, essentially, just like most of the newest generation models, it thinks, [and] First it plans a strategy for picking up. “

This entire system works in coordination with what Kiela calls ‘the best re-ranker in the world’, which helps to prioritize the most relevant information before it is sent to the grounded language model.

Beyond Plain Text: Contextual AI now reads graphs and connects to databases

While the newly announced GLM focuses on text generation, the Platform of Contextual AI recently added support for multimodal content, including graphs, diagrams and structured data from popular platforms such as such as such as such as such as such as Bigquery” Snowflake” Reds And Postgress.

“The most challenging problems in companies are at the intersection of unstructured and structured data,” Kiela noted. “What I am especially enthusiastic about is this intersection of structured and unstructured data. Most of the really exciting problems in large companies are Smack Bang at the intersection of structured and unstructured, where you have some database records, some transactions, perhaps some policy documents, perhaps a lot of other things. “

The platform already supports a variety of complex visualizations, including circuit diagrams in the semiconductor industry, according to Kiela.

Contextual AI’s future plans: creating more reliable tools for daily affairs

Contextual AI is planning to release its specialized RE-Ranker component shortly after the GLM launch, followed by extensive document funeral options. The company also has experimental functions for more agent possibilities in development.

Founded in 2023 by Kiela and AmanPreet SinghHe previously worked at the fundamental AI Research (Fair) team of Meta and Hugging Face, Contextual AI has protected customers, including HSBC, Qualcomm and the economist. The company positions itself as helping companies that finally realize concrete returns on their AI investments.

“This is really a chance for companies that might be under pressure to deliver ROI from AI to look at more specialized solutions that actually solve their problems,” said Kiela. “And part of it is really a well -founded language model that might be a bit more boring than a standard language model, but it is really good to ensure that it is based in the context and that you can really trust that it does its job.”