Models News & Updates

Mistral has launched Devstral 2, a 123-billion parameter dense transformer optimized for agentic software development, along with a smaller variant, Devstral Small 2, and a command-line interface, Vibe CLI. Devstral 2 achieves 72.2% on SWE-bench Verified and is available for free for a limited time via Mistral's API and Hugging Face.

Gemini 3 Pro is Google's most capable multimodal model, delivering state-of-the-art performance across document, spatial, screen, and video understanding, and enabling applications in various fields such as education, medical, and finance.

Z.ai has released an open-source vision-language model series, GLM-4.6V, for multimodal reasoning and frontend automation. The model achieves state-of-the-art results across various benchmarks, including general VQA, chart understanding, and STEM reasoning.

Gemini CLI is a locally installed tool allowing users to access AI without a web browser or GUI, providing a powerful productivity boost for Linux users.

OpenAI's GPT-5.2 update is set for early release next week, aiming to close the gap with Google's Gemini 3 model.

Gemini 3 is a state-of-the-art multimodal AI model that can learn, build, and plan across various domains, offering a major leap in productivity and versatility.

NVIDIA's Kling 2.6 model is now available on the fal platform, enabling native audio generation for text-to-video and image-to-video tasks.

Claude Code now supports fine-tuning of open-source large language models (LLMs) using Hugging Face skills, enabling users to fine-tune models with ease and speed up the development process.