Topic: Models
219 stories · page 1 of 5
New language, multimodal and open-weight models: GPT, Claude, Gemini, Llama, Qwen, DeepSeek and dozens of smaller releases. Each story covers what the model is, how it differs from the previous version, and where to get it.
Weights, licences and hardware requirements go here too.
- Liquid AI releases LFM3, a 2B-class family of on-device multimodal models
- Hy-Nomas: 8B video world model takes free-form text and camera control
- Google DeepMind announces Genie 4, a real-time world model with native audio
- OpenAI launches GPT-6.1 Terra at $1 per million input tokens
- a16z.news essay: superhuman AI hardware engineers are coming, but the physical world will slow them
- WeVisOne-VL unifies object and pixel perception in one open vision-language model
- Kimi releases K3.1 Pro, a 3T-class open-weight model with 1M context
- IQuest-Q1 released as open 320B MoE model for agentic coding
- Block presents Dessa, a full-duplex speech-to-speech model
- OlmoEarth v1.3 releases open 10 m satellite embeddings for the whole planet
- IBM releases Granite Timeseries Ensemble of four forecasters
- Hugging Face releases Carbon-X, an open model that lets you chat with DNA
- Canva releases Canva Craft, its first in-house image generation model
- Recursive releases Flint, a free laptop coding model it says AI largely built
- NVIDIA introduces Kumo Relational, a no-training model for relational databases
- Inception's Mercury 3 diffusion model claims to beat GPT-6 Luna and Claude Haiku 5.5 on speed
- IBM ships Granite Speech 5.0 2B, #2 open-weight model on English ASR
- Sage-A1.5 report describes a 35B MoE agent that delegates and verifies
- NVIDIA releases Cosmos3-H-Dreams, a real-time generative surgical simulator
- Mistral releases Voxtral 3 speech models and a voice agent SDK
- Elevenlabs releases Eleven Realtime, two speech-to-speech models for live conversations
- OpenAI says it is improving ChatGPT for low-resource programming languages
- MiniMax launches M3.1 Flash, says usage costs fall 70%
- Kling AI releases Kling 4.0 with stereo audio, multilingual lip sync and a Flash variant
- Apertus 1.5 adds image and audio input, thinking mode, 262k context
- DeepSeek releases DeepSeek-V4.1-Flash-1017 under MIT License
- Ultramarine 1 Flash claims to beat pricier models on its own benchmark
- Reflection releases Beam, a 501B open MoE model with 1M context
- Jev reaches 35% of Vercel's paid teams in two weeks, Vercel data shows
- Elorian releases Glimpse, a model that re-inspects images while it reasons
- Nvidia releases open Nemotron 3.5 Super VL, a 120B model for enterprise agents
- Meta releases Muse Spark Open, an 816B-parameter open-weight model
- Granite Switch 4.1 bundles chat, decisions and speech in one model
- Google DeepMind unveils Gemini 4 Argon, a physical reasoning model for robotics
- FLUX.3 Image Klein arrives in Diffusers with NVFP4 and FP8 weights
- TypeSafe releases Jev 1.14.0 with opt-in image, box and long-context flags
- OpenAI reportedly plans one or two new models this year, but its flagship still needs alignment work
- LensErase: mask-based erase model trained on 101M examples
- Open vs closed AI models: benefits and risks after Anthropic's GLM-5.3 warning
- Liquid AI releases LFM2.5-24B-A2B, an open-weight 2B-active reasoning agent model
- Google ships Gemini 4 Argon with a 1M-token output window
- Meta ships Muse Wonder and Muse Wonder Full agentic models
- Odyssey opens Odyssey-3 world model to the public with a free demo
- Google releases EmbeddingGemma 2, an open 740M multimodal embedding model
- Microsoft releases Decision-1, a Qwen3.5-9B model for fast structured decisions
- Essay argues AI refusal is far from foolproof and could enable repression
- Whistle speech-to-text model fits in 16.9 MB and runs on the CPU
- TII releases Falcon-ASR, a 1.6B Arabic speech model with Emirati focus
- StepFun's Step 5 Preview, a 600B MoE for agentic work, appears on OpenRouter
- Developer argues DeepSeek 4.1 Flash works like a frontier model at a fraction of the cost