Models
The latest AI foundation models: GPT, Claude, Gemini, Llama, Mistral, DeepSeek, Qwen… Announcements, benchmarks, capabilities, context windows and comparisons.
Models Jev, the new AI model from a ChatGPT co-inventor, appeals to developers
Jev, a new type of AI model designed by a ChatGPT co-inventor, promises developers a cheaper and faster path to software.
Read the article
Models Gemini 3.8 Live: Google combines real-time voice and deeper reasoning
Google DeepMind launches Gemini 3.8 Live and its Extended Thinking version, combining real-time exchanges and deeper reasoning in its AI products.
Read the article
Models AlphaGenome Atlas: Google maps 9 billion mutations
Google DeepMind unveils AlphaGenome Atlas, an AI map of 9 billion DNA mutations to better understand diseases and accelerate research.
Read the article
Models WeatherNext 3: Google makes AI weather forecasting more accurate
Google’s WeatherNext 3 promises more accurate global weather forecasts, particularly for rain and snow, and speeds up weather AI.
Read the article
Models Gemini 3.8 Flash: Google boosts cybersecurity AI
Google unveils Gemini 3.8 Flash and its Cyber version, a more ambitious model for complex tasks, but potentially more expensive.
Read the article
Models Gemini learns to explore videos like an agent
Google DeepMind equips Gemini with agentic video understanding: the AI can explore, search and reason through long sequences.
Read the article
Models Gemini Omni 1.1 Flash gives AI developers more control
Google DeepMind launches Gemini Omni 1.1 Flash, an update focused on developer control to create more reliable AI experiences.
Read the article
Models Ox Alpha: the mysterious AI model setting the web abuzz
An AI model called Ox Alpha is attracting attention without revealing its creators. What is known, the hypotheses, and the stakes of this discreet release.
Read the article
Models Gemini 3.7 Flash: Google further accelerates the AI race
Google DeepMind unveils Gemini 3.7 Flash, a new fast variant in its lineup, expected for real-time use cases and AI agents.
Read the article
Models Meta doubles the efficiency of its advertising AI model
Meta says it has doubled the training efficiency of GEM, its advertising recommendation model, without sacrificing performance at scale.
Read the article
Models Alibaba launches its most powerful model to challenge the United States
Alibaba says its new model, the group’s most ambitious, rivals the best American AI systems. A strong signal for China.
Read the article
Models Gemini 2.5 Pro: Google bets on multi-step agents to win over enterprise
Google DeepMind launches Gemini 2.5 Pro with advanced agentic capabilities and native integration into Workspace.
Read the article
Models GPT-5: OpenAI prepares a native multimodal model for this summer
Sam Altman confirms that GPT-5 will natively integrate vision, audio, and reasoning into a single unified model.
Read the article
Models Claude 4: Anthropic pushes the boundaries of reasoning with its new model
Anthropic unveils Claude 4, a model with advanced reasoning capabilities and a 1 million-token context window. A major step forward for autonomous AI agents.
Read the article
Models Claude Opus 5: Anthropic democratizes its flagship model
Anthropic launches Claude Opus 5, a model close to Fable 5 but cheaper and less restrictive, reigniting competition among premium AIs.
Read the article
Models GPT-5.5 Instant becomes ChatGPT's new engine
OpenAI is rolling out GPT-5.5 Instant as ChatGPT's default model, with fewer hallucinations and reduced latency for sensitive use cases.
Read the article
Models Gemini 3.6 Flash: Google boosts speed and cybersecurity
Google launches Gemini 3.6 Flash, 3.5 Flash-Lite and Flash Cyber: three models designed to accelerate AI use cases, including vulnerability detection.
Read the article
Models OpenAI says it solved an 80-year-old math problem
OpenAI claims its reasoning model solved a geometry conjecture open since 1946, with validation from mathematicians.
Read the article
Models Google unveils an “anything-to-anything” AI model
Google is presenting a multimodal model capable of converting almost any type of input into varied outputs, a key step toward universal creative AI.
Read the article
Models Microsoft launches MAI-Thinking-1, its first reasoning model
With MAI-Thinking-1, Microsoft enters the race for advanced reasoning models and strengthens its in-house AI strategy against OpenAI and Google.
Read the article
Models GPT-Rosalind: OpenAI sets its sights on biomedical research
OpenAI introduces GPT-Rosalind, a model focused on life sciences, medicinal chemistry, and genomics to accelerate biomedical research.
Read the article
Models Claude Fable 5: Anthropic finally opens Mythos to the public
Anthropic is launching Claude Fable 5, the first public version of its Mythos range, with game generation and stronger guardrails.
Read the article
Models GPT-5.6: OpenAI launches its new family of models
OpenAI has officially unveiled GPT-5.6, a new family of models with stronger cybersecurity capabilities and designed to power Copilot and professional use cases.
Read the article
Models Grok 4.5: xAI wants to establish itself among premium models
xAI unveils Grok 4.5, a model presented as “Opus-class,” cheaper and more efficient, reigniting the battle over high-end LLMs.
Read the article
Models GPT-5.6 Sol: OpenAI unveils a model under high tension
OpenAI introduces GPT-5.6 Sol, a model stronger in code, science, and cybersecurity, amid unprecedented restrictions imposed by Washington.
Read the article
Models GPT-5.6 launches under political pressure in Washington
OpenAI unveils GPT-5.6 in a climate of tension with Washington, amid increased capabilities, stronger safety, and a limited rollout.
Read the article