150 results for “TPU”
Google will launch a TPU-equipped satellite on a SpaceX Falcon 9 on October 1 to test AI hardware in orbit. The mission kicks off Project Suncatcher, which aims to run AI workloads from satellites.
SenseTime fully open-sourced SenseNova U1, a native unified multimodal model that fuses vision and language without a vision encoder or VAE. The 8B-parameter release matches commercial models on infographic generation and adds native 4K output in its U1.5 update.
Microsoft Foundry added structured outputs, web search, web fetch, MCP connector, and tool search to Azure-hosted Claude models. The update lets regulated enterprises run agentic workflows without moving data outside their Azure subscription.
After an August 6 warning of a “significant” API increase without numbers, DeepSeek published concrete V4-Flash and V4-Pro peak/off-peak rates effective 16:00 UTC on 16 August 2026 — peak output roughly four times the old flat list.
A research-stage arXiv paper tests whether the factors LLMs cite for their decisions match measured behavioural influence. Across eight Claude, GPT, and Gemini models, cited top-three factors often missed the strongest drivers.
Tencent previewed Hy Image 3.5, a single model for text-to-image and multi-turn editing under one Tencent Cloud API. It costs $0.024 per 2K image, and its claimed 30% quality gain rests on internal tests alone.
Alibaba's Qwen team open-sourced Qwen-Image-2.1, a 7B model unifying image generation, editing, and transparent RGBA output. Day-0 support covers Diffusers, ComfyUI, vLLM-Omni, and SGLang.
Anthropic released Claude Fable 5.1 for general availability and Mythos 5.1 through trusted access. Benchmarks jump on Terminal-Bench-Science and agentic coding; cache-read pricing drops 75% for roughly 25–45% lower total cost at unchanged $10/$50 input/output rates.
NVIDIA took a minority stake in data-center site developer Cloverleaf Infrastructure to accelerate AI factory construction across the U.S. The partnership integrates NVIDIA's DSX platform early in site design to maximize compute output per megawatt.
MiniMax released Music 3.0, an open-weight model that generates complete songs up to five minutes with structured vocals and evolving arrangements. The model combines an 8B Global LLM with a continuous Flow Matching synthesis pipeline for 32 kHz stereo output.
Princeton researchers shadow-evaluated an AI scientist on two NeurIPS directions; original authors scored the outputs 2/6 and 1/6. Engineering loops worked, research judgment did not — while Anthropic's own data shows steep gains in code execution, not goal selection.
OpenAI opened a preview of Ultrafast, an API tier that runs GPT-5.6 Sol up to 14 times faster on Cerebras wafer-scale chips. The mode targets enterprise workloads like incident response and financial analysis with up to 750 output tokens per second.