Recherche
Agent de programmation d'OpenAI, dans ChatGPT, le terminal et l'IDE
API d'OpenAI pour intégrer les modèles GPT, image, voix et temps réel
Le podcast officiel d'OpenAI
Communauté autour d'OpenAI, ChatGPT et Sora
Exemples et guides officiels pour tirer parti de l'API OpenAI : RAG, outils, embeddings, agents.
Codex CLI, l'agent de code open source d'OpenAI qui tourne dans le terminal, lit le projet, modifie les fichiers et exécute des commandes.
Modèle de reconnaissance vocale d'OpenAI, multilingue et robuste, devenu le standard ouvert de la transcription.
Interface de chat auto-hébergée et complète pour Ollama et les API compatibles OpenAI : RAG, outils, multi-utilisateurs.
Plateforme d'agents de développement (ex-OpenDevin) qui écrivent du code, lancent des commandes et naviguent sur le web dans un bac à sable.
SDK léger d'OpenAI pour créer des agents avec outils, transferts entre agents, garde-fous et traçage. Fonctionne aussi avec d'autres modèles.
Agent de code open source pour le terminal, compatible avec la plupart des fournisseurs de modèles, y compris locaux.
Modèle dont les paramètres entraînés, appelés poids, sont publiés et téléchargeables, ce qui permet de l’exécuter, de l’héberger et de l’adapter soi-même, selon les conditions de sa licence.
Système d’IA publié avec les libertés de l’utiliser, de l’étudier, de le modifier et de le partager, ce qui suppose l’accès aux poids, au code et à des informations suffisantes sur les données d’entraînement.
Open Interpreter, désormais un agent de code en Rust pensé pour les modèles ouverts, qui exécute du code sur votre machine.
Le modèle d'OpenAI qui relie images et texte (2021), brique de base de la génération d'images et de la recherche visuelle.
Extrait le contenu de PDF complexes (formules, tableaux, colonnes) en Markdown ou JSON exploitable par un LLM.
gpt-oss-120b et gpt-oss-20b, les modèles aux poids ouverts d'OpenAI (licence Apache 2.0), avec implémentations de référence.
Framework d'OpenAI pour évaluer des LLM et des systèmes à base de LLM, avec un registre de benchmarks.
API unifiée donnant accès à des centaines de modèles d'IA de divers fournisseurs
Le code de GPT-2 (2019), dont la publication progressive avait lancé le débat sur la diffusion des modèles. Archivé.
Reproduction entièrement ouverte de la recette d'entraînement de DeepSeek-R1, par Hugging Face.
Crawler web open source qui produit un Markdown propre pour les LLM, à lancer localement ou en serveur.
Passerelle qui appelle plus de 100 fournisseurs de LLM avec le format de l'API OpenAI, avec suivi des coûts, quotas et bascule automatique.
Agent open source de Google qui amène Gemini dans le terminal, avec un quota gratuit généreux pour les comptes Google.
Clone amélioré de ChatGPT, auto-hébergeable, qui réunit OpenAI, Anthropic, Google, Mistral et modèles locaux dans une seule interface.
Plus de 100 applications d'agents et de RAG open source à étudier et à réutiliser.
Smartphone agents don't get better by processing more data. They improve by training on as many different apps as possible. The researchers behind OpenMobile-2 built an open training environment with
Télécharge et lance des modèles ouverts en local avec une seule commande, et expose une API compatible OpenAI. Le moyen le plus simple de démarrer avec un LLM sur sa machine.
Cours de Microsoft en 21 leçons pour apprendre à construire des applications d'IA générative, traduit en français.
Nvidia's new data engine Flora labels one hour of robot video for 72 cents instead of $88 with GPT-6 Astra, while scoring slightly higher on quality. The open system, built on three fine-tuned Qwen3.5
👉 AI Agents Studio : https://www.youtube.com/channel/UCAawqobkJZ28OLcYcMgqYaw?sub_confirmation=1 "This video covers the top trending open-source GitHub projects this week. You will see tools for devel
Microsoft Research open-sourced Agensh, a multi-agent harness with no central orchestrator. Self-organizing workers share a Git repo, chat and a findings board, scaling from 1 to 1,024 agents on Progr
Moonshot AI's first visible overseas sales hire marks a new stage in how Chinese developers of open-weight models compete outside China. They have given away model weights to win developers and earned
Plateforme de Comet pour tracer, évaluer et surveiller applications LLM, RAG et agents.
Detect and redact Aadhaar, PAN and UPI data in Hindi, Hinglish and 11 more languages, then classify the text with your own labels. Two Apache-2.0 models that run locally. Source: Image by the author.
Cognition released the Devin Vulnerability Fixes plugin, 3 MIT-licensed agent skills that inventory, patch and test PHP, Node.js and self-managed GitLab deployments hit by this week’s critical RCE fla
Assistant et agent de code open source pour VS Code et JetBrains, configurable avec n'importe quel modèle.
For years, enterprise AI strategy has been dependent on a small set of proprietary, closed model providers, a limitation that has carried cost, lock-in and opacity. Now, open-weight models from Meta,
On est presque tous passés sur Hermes… mais OpenClaw lance la version 2.0, avec le retour de l'abonnement Claude. Est-ce que ça vaut le coup de revenir sur OpenClaw ? J'ai installé les deux agents sur
L'un des premiers agents autonomes populaires (2023), devenu une plateforme pour créer et faire tourner des agents en continu.
Bibliothèque Python qui permet à un agent IA de piloter un vrai navigateur web : cliquer, remplir des formulaires, extraire des données.
Outil Python de Microsoft qui convertit Word, Excel, PowerPoint, PDF et bien d'autres formats en Markdown pour les LLM.
You.com launched Inference Router, an OpenAI-compatible API for 12 open-weight models with a 99.9% uptime SLA, zero data retention, 100% US hosting, server-side web search and no platform fee.
Sharing the latest news and advancements in AI Discover More: 🛠️ Explore AI Tools & News: https://futuretools.io/ 📰 Weekly Newsletter: https://futuretools.io/newsletter Socials: ❌ Twiter/X: https://x.
You can use Photoshop-like keyboard shortcuts to go fullscreen, recover a selection, etc. Also, this version provides image-to-image for Krea 2. Download at https://github.com/SamuelTallet/ZPix Hope y
This video tests Kimi K3.1 Pro, capable 3T-class open-weight model. 🔥 Buy Me a Coffee to support the channel: https://ko-fi.com/fahdmirza #kimik31pro #kimi #moonshotai #k31 #k31pro PLEASE FOLLOW ME: ▶
What if an app had no source code, and every screen was generated the moment you clicked it? I built an app launcher that does exactly that with OpenUI, the open-source spec for generative UI that str
TL;DR: When Coqui AI shut down, they left XTTS v2 without official training code for the DVAE audio tokenizer and Latent HiFi-GAN vocoder, and fine-tuning new scripts (Arabic, Indic, etc.) broke on fi
A research team involving miHoYo’s Honkai AI researchers and university collaborators has open-sourced GameScaling, a family of models that play games by generating keyboard and mouse actions from vis
AI company Anthropic has launched OSS Scanner, a free, opt-in service to find vulnerabilities in open-source projects. Enrolled maintainers receive periodic security reports generated by the company's
The Agentic AI Foundation, a consortium that maintains open-source artificial intelligence tools, today took four new projects under its wing. The technologies were donated by OpenAI Group PBC, Anthro
Joseph Kim leads ChatGPT for Financial Services, which launched a month before this talk with design partners Morgan Stanley and Evercore. He demos the product and explains the "capability overhang":
Chinese artificial intelligence model developers DeepSeek and MiniMax have completed export registration procedures and proceeded with open-source releases and overseas licensing activities, according
Calling all Android users: Blurt is coming to Android! No subscriptions. No trackers. Just fast, accurate dictation, powered by AssemblyAI. Get your FREE API key now and be ready to dictate your texts
OpenAI on Thursday announced that it is donating the Open Responses API specification to the open source Agentic AI Foundation
Check out ThinkRail: https://tnkrl.ai/hey1v1 #sponsored 💻️ Need some help with a project or some consulting? Contact me here: https://www.neuralnine.com/services 🐍 The Python Bible Book: https://www.n
OpenAI has opened advertising on ChatGPT to businesses in Nigeria and South Africa, creating a new channel for brands to
Remplaçant auto-hébergé de l'API OpenAI : texte, image, voix et embeddings avec des modèles ouverts, sans GPU obligatoire.
Gradium has open-sourced Phonon, a family of on-device text-to-speech models. The English Phonon 31m ships under Apache 2.0, fits in 64.5 MB, and starts speaking in about 7 ms on a laptop CPU.
AI leaders want their models everywhere and under their control. They may not get both.
Join this channel to get access to perks: https://www.youtube.com/channel/@aicodeking/join In this video, I'll test Recursive Flint, a new open mixture-of-experts coding model with 40 billion paramete
Vercel released durable-harness, an open-source package that logs every model and tool call so Pi, OpenCode, Claude Code and Codex sessions resume after crashes, timeouts and redeploys on Vercel, Post
Hannes joins us to talk about OpenClaw, a local, open-source AI assistant. Come bring your questions on AI assistants, managing large OSS projects, and of course lobster facts 🦞
Guide de référence sur le prompt engineering, le context engineering, le RAG et les agents, avec articles et notebooks.
Chinese artificial intelligence start-up DeepSeek has fully open-sourced an internally developed framework designed to run AI models efficiently, as a senior researcher at the company described the gr
Moteur RAG complet centré sur la compréhension fine des documents (tableaux, mises en page), avec interface et agents.
Fine-tuning de LLM deux fois plus rapide et avec beaucoup moins de mémoire, avec notebooks gratuits et désormais une interface locale.
Google's Harness² lets AI agents improve their prompts, tools, memory, and scripts through two recursive loops without touching model weights. Learn more, including how to run the free demo yourself.
Explore OpenRouter: https://openrouter.plug.dev/Qq0OWea A free Adobe alternative you can deploy in one click, and Claude runs it for you. Six open-source apps (PhotoCraft, LightCraft, FilmCraft, Effec
Recursive has released Flint, an Apache 2.0 local coding model with 40B total and 2.5B active parameters. It reports 74.2% on SWE-bench Verified and 125+ tokens per second on an M4 Pro MacBook Pro.
The internet is in peril.
Chinese artificial intelligence startup Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co. Ltd. today open-sourced DeepInfer, the inference framework behind its own production ser
👉 Try JetBrains Air in IDEs : https://plugins.jetbrains.com/plugin/33314-air-alpha/ "This video covers the top trending open-source GitHub projects this week, picked for practical developer productivi
Obtenir des sorties structurées et validées (modèles Pydantic) de n'importe quel LLM, avec relances automatiques en cas d'erreur.
Google and Meta open-sourced TorchTPU, a native PyTorch backend for Cloud TPUs. It runs eager by default, compiles through torch.compile and XLA, and replaces PyTorch/XLA for new workloads.
DeepSeek has open-sourced the inference engine it uses to serve its models in production, a step it said last year
DeepSeek has open-sourced DeepInfer, the Apache 2.0 inference framework that serves DeepSeek-V4.1-Flash in production. It targets NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs, and ships offline infere
DeepSeek released DeepInfer, the inference framework behind its large-scale model services, on October 20. The framework supports Nvidia GPUs and Huawei Ascend processors and provides capabilities inc
API de modèles d'IA open-weight hébergés dans l'UE par OVHcloud
Alternative à ChatGPT qui fonctionne entièrement hors ligne sur l'ordinateur, avec un catalogue de modèles à télécharger.
Set up your own suite of Adobe alternatives using open-source tools. You will learn to deploy the full Craft Suite on Railway and connect Claude to automate your creative workflow. This tutorial walks
Mistral AI is back in the frontier AI race. Mistral Large 4, nicknamed “Le Chonk,” is a 1-trillion-parameter open-weight model, trained in two months on 3,800 GPUs in Mistral’s own European datacenter
OpenAI’s efforts to undercut Anthropic on price are producing some results. For instance, in May, Anthropic models powered 57% of the coding work of hundreds of engineers at Rippling, an HR software f
Portage de Whisper en C/C++ pour transcrire rapidement en local, y compris sur Mac, Raspberry Pi ou mobile.
An OpenAI AI model locked inside a secure, internet-free training environment found a way out, reached the open web and contacted an external chatbot. The company only discovered the breach afterwards
AI agents at OpenAI could have hidden hundreds of their own actions from the OpenAI monitoring system, outside evaluators found. That exploit was one of several flaws the safety nonprofits METR and Re
The distributed inference framework makes serving large models on your own infrastructure reliable and affordable for the agentic era.
OpenAI is poaching Anthropic's customers with steep discounts. Companies like Rippling, Replit, and DataCamp say they've cut AI costs by up to 90 percent after switching. Anthropic's growth has slowed
DeepSeek just dropped V4.1-Flash-1017, and this minor update is shaking the whole AI industry. The new model becomes the #1 open source coder on Code Arena, beats GPT-6 Astra on DeepSWE, and builds fu
👉 Try OctopusX AI FREE : https://octopusx.ai/external/competition/competitionPage/competitionPageIndex/?utm=youtube&utm_target=ManuAGI "This video covers top trending open-source Dev Tool projects thi
OpenAI is now training its most advanced AI models without internet access following a string of rogue hacking incidents. In a blog post on Friday, OpenAI says it built an offline testing environment
Client de bureau qui réunit de nombreux fournisseurs de LLM, assistants prêts à l'emploi et agents.
Meta on Thursday launched Muse Spark Open, a new 816B open model (with 42B active parameters) that the company argues
The three OpenAI safety researchers who were fired, Tomek Korbak, Mikita Balesni, and Jasmine Wang, are publicly pushing back against allegations from their former employer. They deny sharing confiden
Get a jump start on the US trading day on "Bloomberg Open Interest." OpenAI seeks fresh funding as investors reassess the AI trade, while SpaceX explores a push into wireless. UBS’s Nadia Lovell weigh
👉 AI Agents Studio : https://www.youtube.com/channel/UCAawqobkJZ28OLcYcMgqYaw?sub_confirmation=1 "This video covers the top trending open-source GitHub projects this week, picked for practical develop
OpenAI is reportedly seeking to raise up to $30 billion at a reduced valuation as a bridge to a potential stock market listing in 2027, according to the Financial Times. Earlier talks had pointed to a
OpenAI now trains its most powerful models without internet access, after an internal research model exploited a DNS misconfiguration in September and reached the open internet. In controlled tests, A
