Qwen Releases Image-2.1-Turbo: Open Weights Image Generation and Editing in 8 Steps
Qwen released Qwen-Image-2.1-Turbo, an accelerated checkpoint on its 7B visual generation architecture that creates and edits images in just 8 denoising steps w…
Global AI Digest
-
-
Tencent's Hy-MT2: An Offline Open-Source Translator Covering 33 Languages at 440 MB
Tencent's Hy-MT2 is a 440 MB open-source AI translator running fully offline across 33 languages, promoted by the official Tencent AI account.
-
TRL v1.15 Released With Fused LM Head Cutting Peak VRAM by Up to 82%
TRL v1.15 introduces a default fused LM head delivering up to 82% less peak VRAM, 7x longer sequences, and significant DPO/GRPO token capacity gains.
-
Requesty Launches Compliance Page With Per-Request Evidence Reports
Requesty announced a new Compliance page that answers where prompts were hosted, whether data is retained or used for training, and what scanning detected, with…
-
Command Code Hooks Let You Block Unsafe Commands and Enforce Project Rules
Command Code introduces Hooks that run before tools execute, allowing users to block unsafe commands like rm -rf and enforce project-specific rules, with docs a…
-
Anthropic Pauses Claude Startups Program Offers After 21x Application Surge
Anthropic pauses the Claude Team plan and $1,000 API credit offers in its Startups program after receiving 21x more applications in 48 hours than in the prior f…
-
GMI Cloud Extends Free Access to Qwen3.8-Max, Qwen3.8-Flash, and Wan3.0 with Higher Rate Limits
GMI Cloud is extending free access to Qwen3.8-Max, Qwen3.8-Flash, and Wan3.0 for an additional 7 days, with increased rate limits across all three models, along…
-
ExplorationBench: A New Benchmark for Measuring How AI Systems Explore
A new research release introduces ExplorationBench, designed to evaluate how AI systems frame hypotheses, design experiments, and learn from results in open-end…
-
B.AI Platform Surpasses 2.9 Million Registered Users
B.AI announced its registered user base has passed 2,900,011, citing daily token throughput in the trillions and a unified entry point for frontier models.
-
MCP Servers Requiring Custom OAuth App Registration Criticized for Friction
The author points out that some companies like Slack require users to register a custom OAuth app to use their MCP servers, creating a frustrating setup experie…
