Contextual AI
@ContextualAI
Followers
5K
Following
225
Media
71
Statuses
189
The unified context layer for Enterprise AI. Bridging the gap between LLMs and enterprise data.
Bay Area
Joined March 2023
🚀 Our team just released the highest performing and most efficient reranker in the world! Not only that, these models are fully open weight and free to use. Read our launch blog here: Congrats to the team on this incredible achievement @halal_george.
contextual.ai
Excited to share that we trained rerankers at the cost/performance frontier and are open sourcing them!. Contextual AI Reranker v2.🚀 Best performing, most efficient reranker.🤗 Open weights (1B, 2B, 6B).🫡 Instruction-following (including recency-awareness).🌐 Multilingual. 1/4
0
0
4
RT @douwekiela: Everyone wants to work with the company that promises the “sexy” AI agent, but ends up deploying the “boring” workflow in p….
0
5
0
RT @sheshanshag: Performance on standard retrieval benchmarks like BEIR/ MMTEB hasn't correlated with performance on real world retrieval e….
0
7
0
RT @NirDiamantAI: New tutorial added: Building RAG agents with Contextual AI. Just added a new tutorial to my repo that shows how to build….
0
1
0
RT @douwekiela: Everyone is posting about the “95%” Fortune article, but it’s worth reading the underlying MIT report, which contains a ton….
0
10
0
🏆 It's official - Contextual AI is now at the top of the FACTS leaderboard for groundedness, beating out strong competition from Gemini 2.5 Pro and GPT-5!. Congrats to our research team @w33lliam @rajan__vivek @nandita__naik @Thienhn97 @sheshanshag @shikibmehri on this awesome
Tired of seeing O3 hallucinate? 😵💫.Today, I am excited to share how we built the least hallucinatory LLM in the 🌍. Our GLMv2, developed at @ContextualAI, just claimed 1st place 🥇 on the FACTS Grounded leaderboard by Google DeepMind — outperforming Gemini-2.5-pro, Claude 4, and
0
6
17
Thrilled to see ContextualAI's RAG agent included in this excellent resource for engineers building production-ready agents!.
Incredible milestone!. My Agents-Towards-Production GitHub repository just crossed 10,000 stars in only two months! 🌟. Here's what's inside:. ✅ 33 detailed tutorials on building the components needed for production-level agents. ✅ Tutorials organized.
1
0
6
If you can’t measure generated code quality, you can’t improve it. We recently helped a customer who had an extensive, multimodal codebase and quickly discovered that the main hurdle to generating high fidelity code was evaluation at scale. Traditional token and embedding-based
4
1
8
RT @douwekiela: GPT-5 is out. Early results (and pricing) are truly impressive. Despite that, we may see that the pace of adoption (at leas….
0
2
0
Open source AI development is critical because it democratizes access to cutting-edge architectures, enables researchers to experiment with the latest and greatest models, and creates a collaborative ecosystem where community-driven improvements can compound exponentially. While.
Tested the new @OpenAI OSS 20b & 120b models and the result isn't too promising yet. They are nowhere close to o4-mini performance, at least on the multi-hop QA FRAMES benchmark with GoogleWebSearch on OpenAI-Agent-SDK (or maybe I am doing something wrong -- have been following
1
0
4
RT @douwekiela: The future of AI agents isn't just about better models, it's about smarter context engineering. Today's manual approaches (….
0
1
0
RT @NinaLopatina: We had an interesting meta-learning at @aiDotEngineer World’s Fair from some of the organizers of the MCP track: there ha….
0
3
0
RT @douwekiela: Context engineering for static workflows is relatively straightforward: a unidirectional pipeline with pre-determined tools….
0
5
0
RT @douwekiela: Context engineering has become the critical bottleneck for enterprise AI. Your AI agent works perfectly in demos but breaks….
0
7
0
Continuing Open Source Summer with LMUnit! 🚀. Natural language unit testing that dominates the leaderboards:.✅ Top spots on RewardBench2.✅ SoTA on Flask & BigBench. Read the blog or paper, or find the model on @huggingface with the links below! 👇
📢 As promised ✨, we're open-sourcing LMUnit! Our SoTA generative model for fine-grained criteria evaluation of your LLM responses 🎯. ✅ SoTA on Flask & BigGbench.✅ SoTA generative reward model on RewardBench2. 🤗 Models available on @huggingface: 💻.
0
1
7
RT @rajistics: Introducing: Contextual AI MCP Server (now hosted). After great feedback with our local MCP server, we have added a hosted M….
0
9
0
Our research team continues to push the envelope for what is possible! Topping the FACTS leaderboard is a crucial step in our mission to understand and reduce LLM hallucination - building AI systems that actually do what we want them to do. Worth reading the full thread below 👇.
Tired of seeing O3 hallucinate? 😵💫.Today, I am excited to share how we built the least hallucinatory LLM in the 🌍. Our GLMv2, developed at @ContextualAI, just claimed 1st place 🥇 on the FACTS Grounded leaderboard by Google DeepMind — outperforming Gemini-2.5-pro, Claude 4, and
0
3
18
We had a blast meeting everyone at AI Engineer World's Fair last month! If you missed our workshop and want to learn how to build a production-ready RAG agent in 15 minutes, you can now watch the full recording and follow along at your own pace. Even better, the Contextual AI
1
1
2