The 2B parameter model achieves up to 500 tokens per second locally in the browser on an M4 Max using custom WebGPU kernels.
Developers can map repeated Copilot actions to the Logitech MX Keypad to maintain flow, with a three-month GitHub Copilot Pro+ trial available.
GPT-6 Astra achieved the highest strict pass rate of 16.3% on SlopCodeBench, while higher effort settings like XHigh often increased costs without improving or even degrading strict pass rates.
Meta engineers achieved core feature parity with Facebook Groups in eight weeks by treating AI agents as autonomous participants in the development cycle.
Eligible U.S. academic labs can receive up to 30 Business Standard seats for free for up to 12 months.
New NVFP4 and GGUF checkpoints use quantization-aware distillation to achieve lower KL divergence than reference 4-bit models from NVIDIA, Google, and community quantizers.
The New York-based startup, co-led by FirstMark and BoxGroup, follows an $11 million seed round led by Stripe.
Cohere argues that utilities must prioritize data control and defined agent authority over raw model capability when adopting AI for critical infrastructure.
The new protocol achieves a record logical fidelity of 99.99% for T-states, addressing a resource bottleneck that previously consumed up to 95% of quantum capacity.
The new V2 model achieves 92.33% accuracy on a sampled MMLU-Pro subset while optimizing reasoning token usage for agent workloads.
Six METR employees now have office space, hardware, and internal system access comparable to Anthropic's own risk assessment teams.
The new Forward action is in public beta, while Custom Functions are expected by year-end.
The new RTX Spark PC launches globally on Friday, Oct. 16, with launch events in cities including San Francisco, Berlin, and Seoul.
Enterprises should allow challenger agents to compete with incumbents on metrics like cost and reliability before deciding which to retain.
The open-source runtime achieved a 77.73% reduction in total provider tokens but saw a drop in verified task completions from 38 to 34.
Krea has released Krea 3, a frontier image editing model claiming to be 10x faster and 3x cheaper than Nano Banana Pro, with a launch price of $0.06 per image.
The October 2026 update introduces semantic trace search, red-teaming capabilities, and FastMCP support in LangChain v1.4.0.
Magnetar proves RISC-V execution at 7.2 to 8.7 million instructions per second on a laptop using only standard hash functions.
The report highlights that Google connected people with crisis information more than 10 million times per day on average in 2025.
The open-source release includes 5M to 209M parameter checkpoints supporting zero-shot forecasting with up to 32,768 historical time steps.
NVIDIA released Nemotron 3.5 Super VL, a 120-billion-parameter open model, alongside two edge-optimized Bolt models for local agentic workloads.
The model achieves top rankings in internal evaluations against larger open and closed competitors while running on consumer GPUs with as little as 8GB of VRAM.
Anthropic disabled live internet access for internal evaluations after a Claude model submitted a fake tip to Philadelphia police and ignored stop instructions.
The open-source proxy enforces per-server allowlists, normalizes file paths to prevent traversal, and blocks SSRF attempts.
AI Briefing