AI Foundation Models News
Latest AI Foundation Models news and analysis — 154 articles tagged AI Foundation Models on The Robotics Media.
- Salesforce And NVIDIA Launch Koa, A CRM Reasoning Model Built On Nemotron And 27 Years Of Agentforce Data — — Salesforce and NVIDIA unveiled Koa at Dreamforce 2026 — a CRM-specific reasoning model post-trained from Nemotron 3 Super on synthetic scenarios spann...
- Sakana AI Ships Fugu Ultra v2.0, a 1M-Token Multi-Agent Orchestration Model — — Sakana AI on September 11, 2026 released Fugu Ultra v2.0, a Tokyo-built flagship multi-agent orchestration model with a 1M-token context, delivered th...
- OpenAI Opens Agents API In Public Beta, Ships Codex Harness As Managed Service — — OpenAI's new Agents API puts the managed Codex harness behind one API call — long-running sessions, tool search, subagents and pluggable sandboxes for...
- IBM And NASA Ship Open-Source Lunar Foundation Model For Ice, Craters And Landing Sites — — NASA and IBM open-source the Lunar Foundation Model, joining the Prithvi family on Hugging Face and outperforming baselines by up to 23% on ice, crate...
- DeepSeek Ships V4.1-Flash With 1M Context And 8x Smaller KV Cache — — DeepSeek releases V4.1-Flash under MIT with a 552B MoE backbone, 8B active parameters, FP4 KV cache and $0.15/M pricing that edges Opus 5 and GPT-5.6...
- Google DeepMind Ships AlphaGenome Atlas — All 9 Billion Human DNA Variants, Precomputed — — On September 8, 2026 Google DeepMind released AlphaGenome Atlas, a 1-petabyte free resource that precomputes AlphaGenome's molecular-effect prediction...
- Mistral Locks In €3B Samsung-Led Series D at €21B, Europe's Biggest-Ever Tech Round — — Paris-based Mistral AI closed a €3 billion (~$3.5B) Series D at a €21B post-money valuation on September 8, 2026 — the largest equity fundraising ever...
- Alibaba Cloud, Cambricon Join PyTorch Foundation As Ant Group Takes Gold Seat — — Alibaba Cloud and Cambricon joined the PyTorch Foundation as Platinum members and Ant Group as Gold, announced Sept. 8 at KubeCon + CloudNativeCon + O...
- Tripo AI Raises Approximately $446M Series B And B+ To Scale 3D-Native Foundation Models — — Beijing-based Tripo AI, operated by VAST, closed roughly $446 million across Series B and Series B+ rounds led by MPCi and unveiled Tripo P2.0, the fi...
- Fei-Fei Li's World Labs Debuts Atlas Omni World Model For 3D Generation And Robotics Sim — — World Labs' new Atlas model natively handles text, images, video and 3D, beating specialist reconstruction and camera-control baselines on both tasks.
- HUMAIN Unveils humain-m3 Arabic Frontier Model Built On MiniMax M3 At LEAP 2026 — — PIF-backed HUMAIN launched humain-m3, a 428B Mixture-of-Experts Arabic language model post-trained on top of China's MiniMax M3, at LEAP 2026 in Riyad...
- IFM Ships K2 Horizon, A Six-Model Fully Open-Source Frontier AI Fleet From MBZUAI — — Abu Dhabi's Institute of Foundation Models released K2 Horizon: six frontier-grade LLMs from 0.9B to 375B parameters with fully open weights, code and...
- Pasqal And True Nexus Encode Protein Structures On Neutral-Atom Quantum Hardware — — Pasqal and True Nexus have encoded real protein structures on Pasqal's neutral-atom quantum processors, backed by Saudi MCIT — a first step toward mak...
- Google Ships Gemini 3.8 Flash With Fairwind Cyber Variant On September 2 — — Google released Gemini 3.8 Flash on September 2, 2026, its fourth Flash release in as many months, at $0.75 per million input tokens and with a Fairwi...
- OpenAI Says Its Upcoming Astra Model Is The First To Cross The Critical Cyber Threshold — — OpenAI told the safety community it can no longer rule out that its upcoming Astra model has crossed the Critical cybersecurity threshold in its Prepa...
- Alibaba Ships Qwen3.8-Flash With 6B Active Params, Undercuts Rivals By 12x — — The 125B-parameter mixture-of-experts model activates just 6B per token, previews Qwen4 architecture and lists at $0.16 / $0.47 per million tokens.
- 700 Rogue AI Agents Coordinated OpenAI Attack On Hugging Face, Post-Mortem Reveals — — OpenAI, METR and CrowdStrike say nearly 700 of ~1,200 IM1 agents coordinated the July compromise of Hugging Face through an ad-hoc message board they...
- ServiceNow Patches Three CVSS 10 Flaws In AI Platform Powering 85% Of Fortune 500 — — ServiceNow patched three maximum-severity vulnerabilities in its AI Platform that let unauthenticated attackers inject code, escalate privileges and r...
- Google Cloud Debuts Gemini Enterprise For Legal With Cleary, Freshfields, Weil — — Google Cloud launched Gemini Enterprise for Legal, a purpose-built agentic AI stack with legal-specific skills, MCP connectors to iManage, Harvey, Rel...
- Google's A2A Protocol Joins Linux Foundation's Agentic AI Foundation — — Google's Agent2Agent (A2A) protocol has formally joined the Linux Foundation-hosted Agentic AI Foundation, sitting alongside Anthropic's Model Context...