AI — Page 18

Researcher Pardis Sadat Zahraei and collaborators tested 15 models to see whether matched persuasive prompts change decisions differently across demographic groups.

Anthropic's Claude generated 318,484 candidates in a 64-hour campaign; lab tests found binding, not evidence that any design neutralizes venom or treats a sting.

IBM says the CNCF-governed project now supports multi-tenant quality of service, failover and rolling upgrades, building on Carlos Costa's push to make AI serving cloud-native.

DeepInfer targets large-scale serving and supports DeepSeek-V4.1-Flash, but its README says a complete inference service requires components the project does not include.

The company says GPT-6 Astra passed 72.1% of COBOL tests in its SQLite-building evaluation; its separate migration results use OpenAI-designed hidden tests.

Co-founder Ling Yang's DiscoveryRSI trains research agents using evidence and revision histories; its literature-match result does not show that researchers used the ideas.

The 1-bit GGUF requires at least 128GB of memory, and Unsloth's 88.2% figure measures token agreement on one test set, not general benchmark accuracy.

The Ohio cluster's first phase trained Muse Wonder; Meta says firmware fixes have since raised its interruption interval beyond 54,000 GPU-hours.

Founder Sam Pasupalak says the model plans engineered objects in parts. Skyfall claims large speed, cost and accuracy gains; access is currently offered by waitlist.

The service lets developers set the training loop and keep the resulting weights, while Together supplies tuned GPU infrastructure and deployment.

Co-founder Akshay Kothari says new infrastructure raises the task limit from five minutes. Business and Enterprise seats now include monthly Standard and Select credits; admins can buy more, while Notion says its AI Command Center and spending limits are coming soon.

The connector brings listings, maps and affordability tools into Claude, while keeping Zillow's tour, agent and financing pathways in the conversation.

Arena's October 19th snapshot ranked the October 17th model update No. 11 overall, based on 7,498 agent sessions.

AA-PersonalityBench ranks 30 writing traits and measures 21 style patterns, giving developers another way to compare models for different tasks.

The comparison covers Muse's first 36 days after its September 8th launch; earlier third-party estimates also put the app ahead in its opening 12 days.

A comparison of preserved web-client bundles shows the organization setting appeared between October 14th and October 19th. The code does not establish a customer rollout or reveal how the setting would work.

Founder Andre Watson is steering users to LigandForge and an Anthropic-Adaptyv challenge that will test selected designs for receptor activation.

Madry says he will join Chief Scientist Jakub Pachocki after concluding that increasingly capable AI demands a stronger focus on safety.

The Workers AI preview extracts, calculates and estimates typed values for $0.10 per million input tokens; Reneau previously co-founded Ensemble AI.

Founder Aly Moursy is taking the service from an earlier Egypt-focused beta to a broader offering that prepares applications, checks documents and books appointments.

The integration adds Firecrawl tools to Gemini, Google's Agent Development Kit and Gemini CLI, with enterprise purchases billed through Google Cloud Marketplace.

Tarush Agarwal's benchmark puts Inworld TTS-2 Flash first for full-sentence latency, while Fish Audio Drama 3 leads an automated naturalness estimate.

The free lineup includes DeepSeek, Qwen and other models, but Infron's standard free tier requires a $5 account balance and shares a 1,000-request daily cap.

Founder Erik Newsham's new program bundles three resources for applicants, while its grant size, selection terms and participating labs remain unspecified.