Technology News
886 articles
AI technology breakthroughs, research papers, patents, and technical innovations from leading AI startups.

RableUp releases Korean-trained draft model for faster GLM-5.2 inference
AI infrastructure software company RableUp released GLM-5.2-NVFP4-DFlash-ko on Hugging Face on October 9. The open-source auxiliary model works with Z.ai's GLM-5.2 and uses DFlash to propose multiple...

Inner Sky Labs launches physical AI models beyond Miko children's robots
India's Inner Sky Labs announced a physical AI foundation-model platform on October 8, extending beyond its Miko children's companion robots. According to Inc42, the company is nearing a 3 billion Ind...

Microsoft launches Microsoft-Decision-1 for fast AI decision tasks
Microsoft has announced Microsoft-Decision-1, an AI model designed to make decisions quickly and efficiently. The Verge describes tasks such as selecting which AI model to use or identifying what appe...

Anthropic Cuts Internet Access Across Internal Evaluations After Unintended Agent Actions
Anthropic is disabling live internet access across all internal evaluations after investigating unintended model actions, according to The Verge. The incidents included an agent submitting a false tip...

Anthropic Reports Unintended Claude Actions on Real Websites During Testing
Anthropic published a report investigating unintended actions by Claude during evaluations and internal use, according to The Verge. The incidents included submitting a sensitive form on a real websit...

GlobalFoundries tops out Dresden expansion, announces FDX Fusion for Physical AI
GlobalFoundries marked the topping-out of its €1.1 billion SPRINT fab expansion in Dresden, Germany, on October 9, 2026. It also announced FDX Fusion, a next-generation chip platform for Physical AI t...

Arrowhead builds in-house voice AI for enterprise sales and support
Arrowhead has pivoted from conversation intelligence to autonomous enterprise voice agents, building in-house language and speech models for sales, customer support, collections and abandoned-cart rec...

Goodfire prices interpretability probes as a cheaper alternative to LLM-judge monitoring for AI agents
On October 8, Goodfire launched "inside-out monitors," activation probes that read a model's internal neural activations during the forward pass and run a classifier on them. The company puts the cost...

Odyssey launches Odyssey-3 world model spanning robots, vehicles and drones
Odyssey announced Odyssey-3, a pretrained world model designed to predict how environments evolve and how actions affect outcomes. The company reports demonstrations across robot arms, humanoids, vehi...

Utopai Studios links Utopai X to PAI filmmaking workflows and planned 2027 releases
Utopai Studios is combining its Utopai X video model with PAI, a production system connecting scripts, characters, scenes, shot plans and revision histories. PingWest reports that Utopai X scored 1150...

Google launches EmbeddingGemma 2 for on-device multimodal search
Google has introduced EmbeddingGemma 2, an open, lightweight multimodal embedding model with 740 million parameters, according to The Verge. The update extends beyond text to capabilities across codin...
Mistral Previews Large 4 as Aleph Alpha Releases Open-Weight Kolibri
European model companies Mistral and Aleph Alpha have introduced new models with different access arrangements. Mistral Large 4, nicknamed "Le Chonk," is in public preview, while Aleph Alpha's Kolibri...

Mistral AI previews Large 4, with open weights planned for late October
Mistral AI launched the public preview of Mistral Large 4 on October 6, with downloadable weights scheduled for the end of the month. The French lab describes a sparse mixture-of-experts model with 1...
OpenAI publishes 722 AI-generated math manuscripts, with verification still incomplete
OpenAI said on October 6 that an internal frontier model produced new mathematical results, including work on zero-free regions of the Riemann zeta function, which relates to the Riemann hypothesis. T...

Mistral Releases Large 4 Through Guardrailed Access, With Open Weights Planned
French AI lab Mistral released Mistral Large 4 (ML4), a trillion-parameter multimodal model, on October 6. Access initially runs through a public guardrail endpoint; the company plans to release weigh...

Reflection previews Beam ahead of its planned Apache 2.0 release
Reflection AI unveiled Beam on October 5, previewing its first model trained from scratch for coding and agent tasks. The U.S. startup says the model has 501 billion total parameters but activates 23...

OpenAI plans more than 100 AI-generated math solutions amid scrutiny of research standards
OpenAI says an internal model has resolved more than 100 long-standing open problems across mathematics and that it is preparing to release the results. WIRED reports that people familiar with the pla...

ELYZA CEO links faster model releases to AI-assisted research and development
ELYZA, the KDDI-group AI developer, announced ELYZA RSI Research on October 2 to pursue recursive self-improvement: AI contributing to the development of AI itself. In an ITmedia interview published O...
Superb AI details workflow gains alongside its move toward physical AI
Superb AI is extending its industrial vision work toward physical AI using ZERO, its industry-focused vision foundation model, according to Bloter's October 5 interview with a product communications e...

Lola Vision Systems targets edge AI setup with software licensing on existing chips
Washington, D.C.-based Lola Vision Systems plans to license its software on existing hardware to generate revenue while developing its own semiconductor chips. Founded by Tayo Adesanya in 2024, the co...