IBM is a US technology company that sells enterprise software, mainframe systems and AI products. Recent coverage covered its third-quarter revenue rise of 5% to $17.1 billion, the Granite Speech 5.0 speech models and a Granite Switch 4.1.1 patch, the IBM Z Atlas offering for z/OS, and CEO Arvind Krishna's claim that open weight models deliver 90% of frontier model performance at 10% to 25% of the token cost. IBM also released an open source lunar foundation model with NASA.
Tech leaders adopt Trump's "Super Intelligence" rebrand of AIAfter Trump ordered government documents to replace "AI" with "Super Intelligence" (SI) via executive order, Microsoft, SpaceX, Meta, Tesla and IBM used the term in earnings-season filings. One SpaceX filing used SI 85 times, and Microsoft heralded a "Super Intelligence era."
Paper argues AI is changing the value of a college degreeYilei Shao of East China Normal University argues in an ECNU Review of Education paper that AI making some intelligence abundant undermines traditional human capital theory. Citing Palantir hiring high schoolers and Google and Apple dropping degree requirements, it asks what education should prioritize.
IBM ships Granite Switch 4.1.1 patch releaseIBM released Granite Switch 4.1.1, a backwards-compatible patch that needs no migration. It updates the Speech ASR adapter to handle 10 languages with native punctuation and capitalization, and improves the Decision adapter. Since the base model was untouched, only the adapters were retrained; an Adapter SDK is planned.
IBM launches IBM Z Atlas system intelligence offeringIBM announced IBM Z Atlas, an industry-first Z software offering for z/OS that provides a system intelligence and change orchestration layer, helping mainframe teams assess change impact and cut manual analysis as staff retire and modernization accelerates.
IBM CEO: Open Weight Models Offer Irresistible Value for EnterprisesIBM CEO Arvind Krishna said on the Q3 earnings call that open weight models now deliver 90% of frontier model performance at 10% to 25% of the token cost. He believes most enterprise tasks, such as email summarization and contract review, can use cheaper open models, with only a few requiring expensive frontier models.
IBM Q3 sales up 5% to $17.1B, CEO cites AI tailwindIBM's third-quarter revenue rose 5% to $17.1 billion, beating estimates, with software sales expanding and EPS of $2.95. CEO Arvind Krishna called AI a tailwind for the business.
IBM releases Granite Speech 5.0 speech model familyIBM updated its Granite Speech lineup with the 2B-parameter granite-speech-5.0-2b and two revised models, covering accurate transcription and translation, fast edge transcription and speaker attribution. Trained on 390,000 hours of public audio and supporting six languages, it ships under Apache 2.0 with a 5.06% average English WER, ranking second among open-weight ASR models.
Kyndryl to run AI agents in Toyota Finance production ITKyndryl and Toyota Systems will run AI agents in Toyota Finance's production IT environment for change management, infrastructure design and impact analysis, working from Toyota Finance's own rules and runbooks to flag risks. Contract value, timeline and measured results were not disclosed, but reaching production is the step most such projects never reach.
LeCun Sees Growing Korean Role in World ModelsTuring Award winner Yann LeCun, speaking at the Global AI Leaders Seoul Dialogue, said industrial AI requires "world AI" that understands the physical world and that Korea's role is growing. He noted LLMs' limits in understanding reality and that robots and full self-driving remain unsolved.
Only 10% of IT Chiefs Use Agentic AI for ModernizationA Kyndryl survey of 2,000 senior IT decision-makers found only 10% of IT chiefs apply agentic AI to modernization. About half of projects improved operations, under half saw innovation gains, 18% found limited value, and nearly half ran over budget or schedule. A former IBM strategist says the economics don't hold up.
IBM's Rossi: Trust Is the Real AI ChallengeIBM's Francesca Rossi says security and competitiveness are not opposing goals, and that the real edge will come from trustworthy systems rather than the most powerful models, warning some AI risks are underestimated. The interview follows Dario Amodei's call for a slowdown and Pope Leo XIV's remarks.
NASA and IBM release open source lunar foundation modelNASA and IBM have released the Lunar Foundation Model, one of the first open-source AI models for lunar science. Trained on nearly 2 million tile bundles, mostly from 17 years of Lunar Reconnaissance Orbiter data, it cuts error in predicting polar ice deposits by up to 22 percent versus the strongest model tested against.
IBM and Indian institutes boost AI and quantum researchIBM deepens research partnerships with IIT Bombay and the Indian Institute of Science, focusing on sovereign Indic language model adaptation and autonomous AI agents, building on collaborations started in 2018 and 2021 to advance sovereign AI, agentic AI and quantum computing.
NASA-IBM open-source lunar AI model maps potential water iceIBM and NASA released an open-source Lunar Foundation Model, a multimodal vision transformer trained on geospatial observations of the whole Moon. One key use is locating water ice in permanently shadowed polar regions, a resource for future crewed missions.
Hugging Face releases tokenizers v1 with big speed gainsHugging Face released a v1 release candidate of tokenizers focused on performance, saying it is often tens of times faster than v0.23. The team says tokenization can bottleneck large-scale training, high-concurrency serving and long inputs, leaving GPUs idle. IBM, NVIDIA and the ExecuTorch team contributed patches and cross-hardware testing.
NASA-IBM Lunar Foundation Open-Source Geospatial AI ModelNASA and IBM released Lunar Foundation, an open-source geospatial AI model, but the retrieved text contains only the title and a Hacker News discussion link, lacking technical details or performance data.
NASA and IBM Launch Lunar AI Foundation ModelNASA and IBM launched a foundation AI model for lunar science to support moon research and mission decision-making, indicating the specialization of large-model technology for space-science applications.
IBM releases Granite PatchTST-FM-r2IBM released Granite Time Series PatchTST-FM-r2 (~385M parameters) with an updated architecture, larger pretraining corpus, probabilistic forecasting and imputation; it ranks top among permissively licensed zero-shot replicable models on GIFT‑Eval (and #2 overall), and its weights and code are open under Apache 2.0/OpenMDW 1.0.
The A.I. Daily