Microsoft Foundry Bets on Open Models With Fireworks
Microsoft Azure's Foundry platform now runs Fireworks AI's inference engine, bringing DeepSeek V3.2, Kimi K2.5, and MiniMax M2.5 into enterprise AI under a unified control plane.
Open source
Mistral Small 4
Mistral AI's unified MoE model - 119B total parameters, 6B active per token, 128 experts, 256K context, configurable reasoning, Apache 2.0 license.
Mistral Small 4: 128 Experts, 6B Active, Apache 2.0
Mistral AI releases Small 4 - a 119B MoE with only 6B active parameters, 256K context, configurable reasoning, and Apache 2.0 license. Plus a new NVIDIA partnership to co-develop frontier open models.
Kraken Ships First Crypto CLI Built for AI Agents
Kraken launched an open-source Rust CLI with built-in MCP server that lets AI agents like Claude Code and Codex trade crypto, manage portfolios, and paper-trade against live markets.
AI Scrapers Are DDoSing Self-Hosted GitLab Servers
AI training crawlers from Anthropic, Google, and OVHcloud knocked a French national research institute's GitLab server offline for hours - nobody at CNRS's Institute of Complex Systems could work.
LangChain Releases Deep Agents for Long-Horizon Tasks
LangChain's open-source Deep Agents framework brings planning, subagents, and persistent context to autonomous agents tackling complex multi-step work.
![FLUX.2 [dev]](https://awesomeagents.ai/images/models/flux-2-dev_hu_94ec9e57c8624223.jpg)
FLUX.2 [dev]
Black Forest Labs' 32B open-weight image model - the most powerful open alternative for text-to-image, editing, and multi-reference generation with up to 10 reference images.
![FLUX.2 [klein] 4B](https://awesomeagents.ai/images/models/flux-2-klein-4b_hu_dec233f8cc7116ef.jpg)
FLUX.2 [klein] 4B
Black Forest Labs' fastest open-source image generation model - 4B parameters, Apache 2.0 license, sub-second generation on consumer GPUs with 13GB VRAM.
Shopify CEO Uses AI Agent to Make Liquid 53% Faster
Tobi Lütke ran Karpathy's autoresearch loop against the Liquid templating engine he created 20 years ago, producing 93 commits from 120 experiments that cut parse+render time by 53% and allocations by 61%.
METR: Half of SWE-Bench Passes Fail Real Code Review
METR found maintainers would reject roughly half of AI PRs that pass SWE-bench automated grading, with a 24-point gap that suggests benchmark scores substantially overstate production readiness.
Italian-Legal-BERT
Italian-Legal-BERT is a 110M-parameter domain-adapted BERT model for Italian legal NLP, trained on 3.7GB of court decisions from Italy's National Jurisprudential Archive.
Italian-Legal-BERT Brings Domain-Specific NLP to Italian Law
Researchers at Scuola Superiore Sant'Anna in Pisa built Italian-Legal-BERT, a 110M-parameter model trained on 3.7GB of Italian court decisions that outperforms general Italian BERT on legal NLP tasks.
