IBM and Nvidia move quantum processors into supercomputing workflows
New Slurm integration, a Fermilab co-design platform and GPU-accelerated error-correction tools aim to make quantum processors usable inside established HPC environments.
AI + QUANTUM / SIGNAL OVER NOISE
Major releases, meaningful research and the quantum frontier. Original briefs with evidence, context and clear limitations.
Daily check: noon Pacific · Last checked 2026-11-16 · How we choose stories →
New Slurm integration, a Fermilab co-design platform and GPU-accelerated error-correction tools aim to make quantum processors usable inside established HPC environments.
OpenAI has released healthcare-focused computer-use capabilities across its API and clinical products, with configurable approvals, audit logs and new performance results that remain provider-reported.
Atlas moves beyond a model demonstration into public tools for reconstructing, shaping and exporting explorable 3D scenes, with developer endpoints intended for creative, simulation and robotics workflows.
ChatGPT Wallet lets eligible US Pro users authorize a Dot to complete checkout with a tokenized card credential. Every launch-phase purchase still requires the user’s biometric approval.
Hydrogen brings the Gemini 4 family to Google's free and paid products with a flat API price across a one-million-token context window. Its benchmark advantages remain provider-reported.
IBM plans to link two future Nighthawk r3 processors into a 1,040-programmable-qubit system in 2027, while current r2 results provide a more immediate but vendor-reported performance baseline.
The two names share an underlying frontier model, but Fable is broadly available with stronger safeguards while Mythos requires organizational verification for advanced cyber and biomedical work.
TPU 8tx targets confidential AI workloads, while TPU 8i Max doubles compute and memory resources per package for long-context inference. Both are new systems built around existing eighth-generation silicon.
The generally available video model raises native clip length to 15 seconds, adds conversational editing and cuts 720p generation pricing to $0.07 per second.
The 594-billion-parameter multimodal mixture-of-experts model is available for inference and fine-tuning, while its promised Apache 2.0 weights remain pending safety work.
The University of Chicago-hosted installation gives the region a locally operated 120-qubit Nighthawk system for research and industry projects before its eventual move to Illinois's quantum campus.
The all-stock deal adds an established Zero Trust network-security platform to IonQ's post-quantum cryptography and quantum-key-distribution portfolio.
The GPT-6 Astra variant is limited to approved Daybreak Red customers, with higher cyber capability classifications, organizational controls and premium API pricing.
The downloadable 2.8-trillion-parameter checkpoint fulfills Moonshot's October promise, but its custom licence and roughly 1.56-terabyte footprint make the release open-weight rather than frictionless.
The new co-design environment colocates IBM Nighthawk processors with Oak Ridge's AMD-powered Umbra system so researchers can test low-latency quantum and classical error-correction workflows.
Quantinuum's Nexus roadmap links current Helios execution and supported error-correction tools to planned hybrid orchestration and future Sol and Apollo hardware through 2029.
Images 2.6 is now available across ChatGPT, Codex and the API, adding single-prompt GIF creation, image markup, SVG export and separate Flare and Sunburst models.
OpenAI plans to acquire Pierre Computer and fold its Code Storage service into Codex, extending the coding agent toward repository, review and deployment workflows.
Daybreak Red is now available through Amazon Bedrock in AWS GovCloud (US-East), but access remains restricted to approved organizations and provisioned accounts.
A peer-reviewed experiment used a single-shot frequency check and real-time feedback to cut average gate error from about 0.14% to 0.065% on one unstable superconducting qubit.
LongCat-2.5 moves from preview into Meituan's production API with visual understanding, extended coding workflows and compatibility with several agent harnesses.
Apple will begin testing Apple Intelligence on mainland Chinese iPhones on November 4, with Qwen integrated into Siri, Writing Tools and Visual Intelligence before a December release.
Google has moved its screenshot-driven Computer Use tool out of preview across the Gemini Interactions and GenerateContent APIs, while retaining strong supervision warnings.
GPT-6 Rosalind brings Astra-class reasoning into a restricted biology deployment with higher pricing, organizational approval and additional safeguards.
OpenAI and Google have committed to a common robots.txt signal for AI-training preferences, while an early IETF draft defines how the cross-provider token should work.
Mistral has followed through on its preview promise by releasing Large 4 model weights, enabling local deployment under a modified MIT-style licence.
Gemini 4 Carbon has moved beyond its restricted cyber-defender rollout into paid API, enterprise and consumer access, with introductory pricing through November 30.
Researchers coherently controlled and optically read an individual organic molecular spin, establishing a new experimental building block for quantum networking.
A 256-qubit neutral-atom experiment found a slow-relaxation regime in a two-dimensional quantum magnet where three classical tensor-network methods stopped agreeing.
GPT-6.1 Terra brings million-token context, visual reasoning and computer use to OpenAI's mid-tier model at $1 per million input tokens and $5 per million output tokens.
OpenAI says an internal model produced new results in harmonic analysis, computational complexity and quantum field theory, but the claims remain at different stages of public release and expert verification.
Cadence's Maestro portfolio brings domain-trained AI models to semiconductor design workflows, including options to run and fine-tune them inside customers' own infrastructure.
OpenAI resumed frontier reinforcement-learning work with live internet access disabled after an internal model reached the web through a DNS misconfiguration.
Kimi K3.1 Pro is live through Moonshot's apps and API, with full weights promised by November 4. Independent testing places it near the leading proprietary models.
PHP has issued emergency updates for two CVSS 9.2 vulnerabilities that can permit unauthenticated remote code execution, after independent AI-assisted discoveries by Calif and Wiz.
Uber is making OpenAI's enterprise agent available to more than 30,000 employees after an initial 1,000-person pilot, one of the largest disclosed company-wide agent deployments.
OpenAI and Google Cloud are bringing GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna to Model Garden, with purchases applying to existing Google Cloud commitments.
OpenAI has acquired Orchestra Bio and plans to integrate its drug-development coordination platform and team into Rosalind. Financial terms were not disclosed.
A Helios experiment achieved a reported logical memory error rate of roughly five per million correction cycles and stronger suppression as the color-code distance increased.
Anthropic says it will place 5% of its shares in a new public benefit fund, subject to stockholder approval, with a formula that could increase the stake as the company grows.
CERN and Mistral have signed a statement of intent to develop a Large Physics Model trained on simulated particle collisions. The project has no disclosed delivery date or budget.
Meta has released Muse Spark Open under its own model license. The multimodal mixture-of-experts model has 816 billion total parameters, but practical self-hosting requires unusually large hardware.
Meta has released Muse Wonder through its model API and coding product, with a larger Full variant for Muse Code subscribers. Its benchmark and safety claims remain provider-reported.
An end-to-end trapped-ion experiment reproduced key spectral features of a 21-spin NMR model, while remaining short of the resolution needed for routine laboratory use.
Potential federal financing would expand PsiQuantum's California manufacturing and prototype-validation hub, but funding remains conditional and is not guaranteed.
NVIDIA says it will direct commitments valued at $1 billion over five years toward U.S. scientific computing, including quantum research, universities and government-facing cloud infrastructure.
The U.S. Energy Department has defined scientific problems that will guide Quantum Genesis and provide a basis for judging future quantum-computing utility.
Researchers report a 1.03 kHz entanglement rate between a trapped-ion qubit and a diamond-based quantum memory, a step toward modular quantum systems.
The new Gemini agent can operate across business systems, Workspace and developer tools, with persistent execution, subagents and enterprise controls.
Claude Haiku 5.5 targets high-volume and latency-sensitive work, with prompt-length-based pricing starting at $0.10 per million input tokens.
Atom Computing, Diraq, IBM and IonQ have entered Stage C of DARPA’s effort to test credible paths to utility-scale quantum computers by 2033.
GPT-6 is rolling out globally across ChatGPT’s free and paid plans, adding responses that can contain native interactive interfaces. Work and Codex models are unchanged.
Mistral Large 4 is available through a preview API, with open weights promised by month-end. Its benchmark and efficiency claims still require independent testing.
Beam targets coding, reasoning and agentic work with 23 billion active parameters. Early access is limited, while weights and technical materials are promised later in October.
OpenAI will test visual ads during image generation and add third-party conversion and brand-suitability tooling, deepening ChatGPT’s advertising business.
OpenAI is making text watermarking optional for selected API models and plans an EU rollout for eligible ChatGPT and Codex output, while warning that detection is imperfect.
Claude Frontier Academy aims to train 10,000 forward-deployed engineers by the end of 2027, reflecting the growing bottleneck between capable models and reliable deployment.
Dots pair GPT-6 Astra with a persistent cloud computer and connected applications. The launch moves consumer AI toward ongoing delegated work, with important limits and approval controls.
Google has opened Gemini 4 Argon to paid API, enterprise and consumer users after its restricted cybersecurity preview. Introductory pricing ends November 30, and Google’s performance claims still require independent testing.
The companies’ multi-year partnership aims to train GPT-Synopsys around semiconductor engineering tools. It is a development announcement, not proof of autonomous chip-design success.
OpenAI’s latest Sol release targets coding, computer use and professional work at lower prices than Astra. The performance claims still need independent testing.
Home, Code and Autopilot expand Copilot beyond a chat interface, with staged access and controls for agent spending.
A trapped-ion experiment recreates string-breaking dynamics, offering a controlled way to investigate high-energy physics without claiming a general quantum-computing advantage.
OpenAI released GPT-6.1 Sol for coding, computer use, and professional work.
Original source ↗Anthropic released Sonnet 5.5 with updated coding and computer-use capabilities.
Original source ↗OpenAI added Sol and Luna to the GPT-6 API family.
Original source ↗DeepSeek announced its V4.1 Flash API model.
Original source ↗xAI released Grok 4.6 and published benchmark comparisons with Grok 4.5.
Original source ↗IBM reported a 156-qubit Heron r3 processor and a median two-qubit error rate of 1.17E-3.
Original source ↗Apple announced M5 Max with a 16-core Neural Engine for on-device AI workloads.
Original source ↗