A new scheduling strategy keeps workloads within NVLink domains to avoid 43% performance loss, yielding 27% more usable GPUs and 14.4% higher throughput in pre-training.
The initiative partners with universities and industry to release scientific artifacts in machine-actionable formats, starting with a CMU-hosted climate prediction challenge on November 2.
The 1T parameter model features a 1M token context window and ranks 5th on the Artificial Analysis Intelligence index for open-weight models.
The packages wrap Apple's new SpeechAnalyzer API with NVIDIA and Alibaba models, achieving 1.55% WER on LibriSpeech test-clean, and enable typed decision outputs via six backends including MLX and llama.cpp.
The article shares how PMs use AI to automate repetitive tasks, allowing them to focus on planning and judgment.
The company plans to produce 20,000 RNGD cards this year and more than double that in 2027, with the chip already in mass production and deployed by customers like Samsung SDS and LG Group.
Liquid AI has released six 2B-scale LFM3 models supporting execution on edge devices and multimodal processing.
Campaigns focused on a single feature drove 25% of desktop visits to new accounts, with trials starting August 28-September 12, 2026.
The new suite uses internal analytics and surveys to provide granular visibility into team output, costs, and workflow friction, with data excluded from model training by default.
The preview allows users to send texts, photos, and voice memos to their agents, which can proactively reply with task updates.
The benchmark tests agents building 17 products from scratch, with top models failing on feature intersections and schema errors.
The new model features M5 Pro and M5 Max chips, a 10,000,000:1 contrast ratio display, and starts at $2,899.
The system uses high-fidelity production replay and parallel sub-agents to diagnose bottlenecks and test speculative decoding, quantization, and scheduling changes.
The feature allows users to explore topics layer-by-layer, from microscopic systems to deep space, with audio narration coming soon.
A study on GPT-OSS-20B shows that a fixed generic feedback line yields the same ~3-point performance gain as self-written critiques in multi-step financial reasoning.
Google confirmed a technical gap in its AI Mode that prevented it from providing direct answers for the 2020 presidential election and 2022 Senate races, an issue the company is currently fixing.
Adding an evidence graph to the OpenWiki memory layer improved accuracy by 27% and cut cost per correct answer by 58%.
The new platform lets developers build once to reach users in both Search and the Gemini app, with initial partners including Adobe, Calm, and Peloton.
The model ranks #1 on the Artificial Analysis Intelligence Index v5.0 at 59 and leads in vulnerability detection with 88.2% on SEC-bench Pro.
The harness, which powers GitHub Copilot and VS Code, supports 20+ frontier models and is built by the GitHub engineering team.
The 1B OCR model runs locally on mobile devices while the 2B verifier model checks output accuracy against original images.
GitHub Enterprise customers can now access all Copilot features without a separate subscription, while Copilot Fusion achieves high performance at 40% less time per task.
GitHub reports 225 million developers and 17.9 billion pushes in 2026, with agent-started pull requests increasing from 17% to over 40% of all PRs.
The new architecture uses a Mixture-of-Experts model and fused GPU kernels to deliver near-ranking-level performance at 300x fewer FLOPs than the 2024 system.
AI Briefing