Daily AI News

OpenAI's GPT-5.6 Launches Thursday After U.S. Government Delay

OpenAI is releasing GPT-5.6 after the U.S. government lifted a release ban following additional testing. The model reportedly beats Anthropic's Claude Mythos 5 on coding benchmarks at roughly half the cost, though binding standards for future model approvals remain absent.

Read source →

xAI Releases Grok 4.5, Described by Elon Musk as 'Opus-Class Model'

Elon Musk's xAI has launched Grok 4.5, positioning it as a cheaper, more efficient alternative to other top-tier models. Musk describes it as an 'Opus-class model,' suggesting performance competitive with Anthropic's highest tier.

Read source →

Mistral Enters Robotics with Robostral Navigate, an 8B Vision-Language-Action Model

Mistral launches its first robotics model, Robostral Navigate, an 8B parameter model that guides robots through unknown environments using only a single RGB camera. Trained in simulation and refined with reinforcement learning (CISPO), it achieves 76.6% on the R2R-CE benchmark.

Read source →

Anthropic's Claude Fable 5 Tops Industry Benchmarks at Steep Cost Premium

Claude Fable 5 sweeps all six new industry-specific benchmarks from Artificial Analysis (finance, law, medicine), but single tasks cost $3.48—over 100x cheaper alternatives like DeepSeek V4 Pro. Anthropic recommends using Fable 5 as a planner delegating to Sonnet 5 for 92% performance at 63% cost.

Read source →

Ant Group's Robbyant Releases LingBot-VLA 2.0: Open-Source 6B Cross-Embodiment Robot Model

Robbyant (Ant Group) open-sources LingBot-VLA 2.0 under Apache-2.0, a 6B vision-language-action model pretrained on ~60,000 hours of data across 20 robot configurations and 10,000 hours of egocentric video. Designed for cross-embodiment manipulation.

Read source →

Chinese Startup MiniMax Plans to Open-Source 2.7 Trillion Parameter Model This Year

MiniMax announces plans to release a 2.7 trillion parameter LLM as open source later in 2026, which would be the largest open-weight model to date by a wide margin.

Read source →

AI Chip Maker SambaNova Raises $1B at $11B Valuation

SambaNova secures $1B in Series F funding at an $11B valuation, just five months after its previous mega-round. The raise comes amid rumors Intel attempted to acquire the company for ~$1.6B.

Read source →

Google AI Studio Adds 'Import from GitHub' to Build Mode for Instant App Deployment

Google AI Studio's Build mode now supports importing GitHub repositories, automatically transforming them into runtime-compatible, editable, and deployable applications—streamlining the developer loop from code to live AI app.

Read source →

LangChain and NVIDIA Launch NemoClaw Deep Agents Blueprint for Governed Enterprise Agents

LangChain and NVIDIA release NemoClaw, a blueprint combining Deep Agents Code, Nemotron 3 Ultra, and OpenShell for open, governed enterprise agents with deny-by-default networking, human approval gates, and audit logs.

Read source →

French Startup ZML Releases Free ZML/LLMD to Speed Inference Across Heterogeneous AI Chips

ZML, backed by Yann LeCun, releases ZML/LLMD—software that optimizes LLM inference across diverse accelerator hardware (Nvidia, AMD, Intel, etc.), potentially reducing costs and vendor lock-in for multi-chip deployments.

Read source →