Sign in

LightOn

@lightonai.bsky.social
117 followers 12 following 279 posts

LightOn is a leading European generative AI company delivering secure on-prem RAG for document intelligence, enabling safe use of sensitive data behind firewall

PostsRepliesMedia
LightOn @lightonai.bsky.social · 28/04/2026
Retrieval is solved. One API to feed any model, power any agent. Three endpoints, zero config: /parse /extract /search LightOn Console dropping soon. Sandbox access, ship as you sign up! 🔗 console.lighton.ai
010
LightOn @lightonai.bsky.social · 25/03/2026
@igorcarron.bsky.social était l'invité de Frédéric Simottel sur BFM Business pour revenir sur les dernières innovations de LightOn et ce nouveau champ qu'elles ouvrent : l'intelligence documentaire.
000
LightOn @lightonai.bsky.social · 25/03/2026
🎙️ "Il faut penser l'IA comme une infrastructure ancrée dans la réalité documentaire des organisations, et non plus comme une application ex machina."
bfmtv.com
Lighton explore les nouveaux champs de l'IA - 24/03
VIDÉO - Ce mardi 24 mars, Igor Carron, PDG et co-fondateur de Lighton, s'est penché sur la place de Lighton dans la chaîne de valeur de l'IA, et la reprise du contrôle sur la recherche avec l'IA, dans...
131
LightOn @lightonai.bsky.social · 24/03/2026
Gauthier Zuppinger brings the technical depth alongside Fabrice Bagniakana. No skipping the hard parts. 📅 March 27 · 11:00–12:00 CET 🔗 Register: events.teams.microsoft.com/event/214067...
000
LightOn @lightonai.bsky.social · 24/03/2026
LightOn is joining @tdsynnex.bsky.social on March 27, to show what production-grade retrieval actually looks like inside a regulated enterprise: 🔍 Hybrid search, 🧠 structured reasoning, 📋 full auditability, 🔒 zero data leaving your infra.
100
LightOn @lightonai.bsky.social · 24/03/2026
To everyone who has hit the wall doing RAG: we planned this one for you. Broken retrieval. Hallucinations at inference. Pipelines that fold the moment data gets sensitive. We know where it breaks. We built Paradigm to fix it.
110
LightOn @lightonai.bsky.social · 19/03/2026
149M parameters. Open weights. Open code. Built with PyLate in a few hours. Full results, analysis and recipe on LightOn blog: lighton.ai/lighton-blog...
lighton.ai
The Bloated Retriever Era Is Over - LightOn
Reason-ModernColBERT tops BrowseComp-Plus, the most rigorous agentic search benchmark, across every metric, with 54× fewer parameters and fewer search calls.
000
LightOn @lightonai.bsky.social · 19/03/2026
Reasoning-intensive retrieval (BRIGHT), code search (MTEB Code), agentic Deep Research (BrowseComp-Plus). The pattern is the same: late interaction dominates, with a fraction of the parameters.
100
LightOn @lightonai.bsky.social · 19/03/2026
The multi-vector era is here and there is no going back. Reason-ModernColBERT tops BrowseComp-Plus, the hardest agentic search benchmark available, by 7.59 points on accuracy. 🥇on accuracy. 🥇on recall. 🥇on calibration. 📉 Fewest search calls. The models it outperforms? Up to 54× larger.
110
LightOn @lightonai.bsky.social · 12/03/2026
🗞️ Read the press release: www.lighton.ai/lighton-blog...
lighton.ai
LightOn Enhances Its Paradigm Platform with Real-Time Web Access Through a Strategic Partnership with Linkup - LightOn
LightOn, a leading French provider of secure AI for sensitive data, announces a strategic partnership with French startup Linkup, which specializes in web search designed for AI applications.
010
LightOn @lightonai.bsky.social · 12/03/2026
Private data and the open web, securely unified in a single pipeline. Together, the two technologies enable organizations to build AI systems that are better informed, more reliable, and designed for demanding environments, combining search and reasoning to deliver accurate and actionable outputs.
110
LightOn @lightonai.bsky.social · 12/03/2026
AI agents are only as good as the information they can access. 🇪🇺 LightOn, an AI Search & Reason company, is partnering with Linkup to provide structured, real-time web search to Paradigm within a secure, fully European technology stack.
110
LightOn @lightonai.bsky.social · 19/02/2026
Models, checkpoints, training code under Apache 2.0. 🧑‍🍳 Kudos to the whole team @nohtow.bsky.social Luca Arnaboldi @amelietabatta.bsky.social @krzakalaf.bsky.social 🔗 Dive into the release: www.lighton.ai/lighton-blog...
lighton.ai
Day Zero of Multi-Vector Retrieval - LightOn
Introducing ColBERT-Zero: late interaction model trained from scratch with PyLate
000
LightOn @lightonai.bsky.social · 19/02/2026
🥇 SOTA on BEIR, <150M params ⚡ Supervised-first → distill = most of the gains for a fraction of the cost 🧠 Prompt alignment is non-negotiable to preserve peak performance through fine-tuning
110
LightOn @lightonai.bsky.social · 19/02/2026
In collaboration with @epfl-ai-center.bsky.social and the Swiss AI initiative, LightOn pre-trained it end-to-end for late-interaction retrieval
100
LightOn @lightonai.bsky.social · 19/02/2026
Day Zero for Multi-Vector Retrieval. Today we’re flipping the retrieval playbook: no dense model adaptation, no retrofit. 🏗️Multi-vector from scratch, powered by PyLate. Meet ColBERT-Zero
110
LightOn @lightonai.bsky.social · 12/02/2026
Give your coding agent the search it deserves. Huge kudos to @nohtow.bsky.social and @raphaelsty.bsky.social Read more: www.lighton.ai/lighton-blog...
lighton.ai
LateOn-Code & ColGrep: LightOn unveils state-of-the-art code retrieval models and code search tooling - LightOn
The "Stronger Grep" for Modern Development While AI coding assistants like Claude Code have transformed how code is written, their ability to navigate large codebases efficiently is often limited by k...
010
LightOn @lightonai.bsky.social · 12/02/2026
What we measured with Claude Code: 🚀 70% win rate vs. vanilla grep 📉 ~60k tokens saved per question 🤏 56% fewer search operations Built in Rust with Next-Plaid - 100% local - No code leaves your machine.
120
LightOn @lightonai.bsky.social · 12/02/2026
ColGrep is powered by LateOn-Code-edge (17M) and LateOn-Code (130M), the first late-interaction models purpose-built for code. 🏆 They top MTEB Code, outperforming models up to 17x their size while running instantly on a laptop.
100
LightOn @lightonai.bsky.social · 12/02/2026
ColGrep mirrors the grep interface your agents already use, but replaces pattern matching with semantic scoring, and supports hybrid queries that combine both. It plugs straight into Claude Code, OpenCode, or Codex.
100
LightOn @lightonai.bsky.social · 12/02/2026
🔥 Stop burning tokens on blind grep searches. Give your coding agent semantic eyes. Meet LateOn-Code & ColGrep: a Rust-powered search tool and two SOTA late-interaction models that bring intent-level code retrieval directly to your terminal.
lighton.ai
LateOn-Code & ColGrep: LightOn unveils state-of-the-art code retrieval models and code search tooling - LightOn
The "Stronger Grep" for Modern Development While AI coding assistants like Claude Code have transformed how code is written, their ability to navigate large codebases efficiently is often limited by k...
150
LightOn @lightonai.bsky.social · 10/02/2026
Huge kudos to @raphaelsty.bsky.social for shipping this breakthrough! 🙌 Read the full article here 👉 www.lighton.ai/lighton-blog...
lighton.ai
Introducing LightOn NextPlaid - LightOn
Multi-Vector Database Built for Sharper Retrieval and Frugal Inference
010
LightOn @lightonai.bsky.social · 10/02/2026
NextPlaid represents the "Blanc" milestone in our Bleu/Blanc/Rouge roadmap for enterprise document intelligence. It follows the "Bleu" release, LightOnOCR-2, a SOTA 1B OCR model which converts complex documents into clean, usable text.
110
LightOn @lightonai.bsky.social · 10/02/2026
⚙️ Production Ready: Built in Rust and optimized for CPUs, it supports incremental index updates and concurrent reads/writes—capabilities missing from standard implementations.
100
LightOn @lightonai.bsky.social · 10/02/2026
🚀 Seamless Integration: NextPlaid runs alongside your existing vector database. You can add multi-vector retrieval to your established RAG pipeline without ripping anything out.
100
LightOn @lightonai.bsky.social · 10/02/2026
📉 Frugal Inference: High-signal context reduces the amount of noise sent to your LLM, allowing it to answer with fewer, more accurate tokens.
100
LightOn @lightonai.bsky.social · 10/02/2026
Why NextPlaid is the missing layer for your RAG stack: 🎯 Precision Matching: Retrieval matches at the token level, surfacing the exact passage that answers your question rather than just a document that vaguely relates.
100
LightOn @lightonai.bsky.social · 10/02/2026
By representing documents as sets of vectors, one per token, we preserve the distinct concepts and precise details that other search engines average away.
100
LightOn @lightonai.bsky.social · 10/02/2026
🔍🪡To find the needle, you better index every straw of the haystack. Today, LightOn is launching LightOn NextPlaid: a CPU-optimized multi-vector database that indexes at the token level.
130
LightOn @lightonai.bsky.social · 09/02/2026
En entreprise : 📄 vos documents sont vivants, 🔍 l’observabilité est indispensable, 🌳 le bruit coûte cher et les GPUs ne poussent pas sur les arbres ! Un épisode dense et sans langue de bois sur l'IA en entreprise. 🎧 Écouter l'épisode 👉 Spotify : open.spotify.com/episode/4Dtt...
open.spotify.com
Comment donner une mémoire fiable aux intelligences artificielles ? avec Amélie Chatelain, Head of Knowledge & Search chez LightOn
010
LightOn @lightonai.bsky.social · 09/02/2026
@amelietabatta.bsky.social Head of Knowledge & Search chez @lightonai.bsky.social est l’invitée de Laurent Nicolas-Guennoc pour le podcast Converteo “Changement d’époque” Face au narratif "bigger context = better", Amélie remet les pendules à l'heure.
110
LightOn @lightonai.bsky.social · 09/02/2026
🎙️“Mettre tous vos documents dans le contexte d'un modèle, c'est comme inviter 30 personnes à une réunion où une seule suffit : ça coûte cher, ça fait du bruit, et au final le résultat est moins précis !”
120
LightOn @lightonai.bsky.social · 28/01/2026
Congrats to @orionweller.bsky.social @jhuclsp.bsky.social @nohtow.bsky.social for pushing the boundaries of useful AI. 🧑‍🍳 Read the open recipe here: lighton.ai/lighton-blog...
lighton.ai
Introducing Ettin Suite: the SoTA open recipe to outperform existing Generative & Retrieval Models - LightOn
Introducing Ettin, the first ever SOTA suite of paired encoder & decoder models, developed by Johns Hopkins University in collaboration with LightOn.
020
LightOn @lightonai.bsky.social · 28/01/2026
Size matters less than the right architecture choice. That’s why the smallest Ettin model is already being massively adopted to build high-performance Edge AI. It’s time to stop forcing "Decoder-only" models on every problem. For high-value tasks, specialized engineering beats generic scale.
130
LightOn @lightonai.bsky.social · 28/01/2026
Ettin was built as the first-ever SOTA suite of paired encoder-only & decoder-only models to prove a point: 🔍 Encoders for classification & retrieval ✏️ Decoders for text generation
110
LightOn @lightonai.bsky.social · 28/01/2026
The Ettin suite paper has been accepted to @iclr-conf.bsky.social It highlights the Elephant in the room: 🏗️ Architecture matters.
120
LightOn @lightonai.bsky.social · 16/01/2026
The "G" in RAG only amplifies what the "R" provides. If your retrieval layer is static, your AI is hallucinating on facts. Here is how LightOn approaches RAG as critical infrastructure, not just a chatbot feature. 👉 www.lighton.ai/lighton-blog...
lighton.ai
RAG isn’t Dead, Yours is! - LightOn
Static ingestion, stale answers, lost trust.
020
LightOn @lightonai.bsky.social · 16/01/2026
When you treat it as a simple add-on: 📉 Relevance drops as document versions change. 🔐 Security blocks you because access control wasn't enforced at query time. ⚠️ Trust erodes because the system generates confident answers based on last week's data.
120
LightOn @lightonai.bsky.social · 16/01/2026
Stop building RAG like a feature. It is infrastructure. RAG inherits every constraint of your organization: scale, heterogeneous data, and strict governance
120
LightOn @lightonai.bsky.social · 15/01/2026
This makes high-performance OCR far more accessible for edge deployments, privacy-sensitive use cases, and cost-efficient production setups. 🔗 Model weights available here huggingface.co/lightonai/Li...
huggingface.co
lightonai/LightOnOCR-1B-1025 · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
010
LightOn @lightonai.bsky.social · 15/01/2026
🚦No GPU required 💻 Runs locally on a laptop (CPU-friendly) 🥇 SOTA performance on your data: Fine-tune easily using standard Hugging Face tooling LoRA, PEFT, Trainer
110
LightOn @lightonai.bsky.social · 15/01/2026
🤗 LightOnOCR-1B is now in Hugging Face Transformers With 1.2B downloads, Transformers Library is the go-to toolkit for developers building AI applications. Any developer can now add state-of-the-art document reading to their app in one line of code. #Transformers #OCR #GenAI
120
LightOn @lightonai.bsky.social · 09/01/2026
By going beyond raw model performance, private RAG provides enterprise-grade AI grounded in internal data, confidentiality constraints, and trusted usage. This is the challenge LightOn addresses. 👉 Read more in our latest post: www.lighton.ai/lighton-blog...
lighton.ai
Beyond Shadow AI: Private RAG as the Foundation of Enterprise AI Adoption - LightOn
The Era of
021
LightOn @lightonai.bsky.social · 09/01/2026
Shadow AI is more than a governance issue. It’s a signal of unmet enterprise AI needs. 🧩 Every prompt shared with public AI systems is a request for performance, context, and usability.
lighton.ai
Beyond Shadow AI: Private RAG as the Foundation of Enterprise AI Adoption - LightOn
The Era of
121
LightOn @lightonai.bsky.social · 09/12/2025
The LLM is a voice. It provides the syntax, the grammar, and the fluency. But the intelligence, the competitive advantage, comes strictly from your data. This is why the future of Enterprise AI will be built on context, not parameters. We break it down here 👇 www.lighton.ai/lighton-blog...
lighton.ai
Why Bigger Models Won’t Win the Enterprise AI Race - LightOn
While the industry races for bigger models, the real revolution for the enterprise is happening in memory, not intelligence. Here is the vision we shared at the Grand Palais.
020
LightOn @lightonai.bsky.social · 09/12/2025
The race for bigger models, is not yours. Your battle is about mobilizing your fragmented, unstructured data to create value!
lighton.ai
Why Bigger Models Won’t Win the Enterprise AI Race - LightOn
While the industry races for bigger models, the real revolution for the enterprise is happening in memory, not intelligence. Here is the vision we shared at the Grand Palais.
120
LightOn @lightonai.bsky.social · 02/12/2025
“Why build RAG pipelines when you could just… dump everything into the context window?” Amélie Chatelain, Ph.D. ran the comparison you’ve always wondered about, long-context versus retrieval, and the results might surprise you. Watch the full talk here: www.lighton.ai/lighton-blog...
lighton.ai
RAG is Dead, Long Live RAG: Retrieval in the Age of Agents - LightOn
A technical deep-dive into why retrieval-augmented generation evolved rather than died, and what intelligent retrieval looks like in 2025.
020
LightOn @lightonai.bsky.social · 25/11/2025
Far beyond accessibility, this collaboration empowers every organizations with the AI capabilities needed to outperform, out-innovate, and out-scale their competitors. 🗞️ Learn more: lighton.ai/lighton-blog...
lighton.ai
LightOn Democratizes Private AI with Validation of Its Platform on NVIDIA RTX PRO 6000 Blackwell Server Edition - LightOn
New compact configuration lowers entry barriers for SMEs, bringing Enterprise Search and Reasoning where operations happen.
010
LightOn @lightonai.bsky.social · 25/11/2025
Why this matters: ⚡ Edge AI first: powerful on-site retrieval & reasoning with deterministic latency 🔐 True sovereignty = your AI, on your infrastructure 🏭 Optimized for factories, hospitals, critical infrastructure & air-gapped sites
110
LightOn @lightonai.bsky.social · 25/11/2025
This milestone makes enterprise-grade AI Search & Reasoning accessible to organizations of all sizes, with compact, frugal configurations that run right where your data and operations live.
100