X Launches Unified Subscription Bundling Grok, Cursor, and X Premium
X announced a single subscription that unifies plans across Grok, Cursor, and X Premium, letting users code with AI in Cursor, use a bot for work tasks, and acc…
Global AI Digest
-
-
Muse Wonder Tested Against Opus 5.5 on 2D Tasks: Cheaper but Less Detailed
A hands-on comparison shows Muse Wonder completing 2D tasks faster and cheaper than Opus 5.5, though Opus still renders details better.
-
Meta's New 'Watermelon' Model Gets Early Access Praise from Reviewer
A reviewer with early access to Meta's new 'Watermelon' model reports strong instruction-following, self-checking coding agent abilities, and benchmark results…
-
Meta's Muse Wonder Tops DeepSWE, Now Available in Cline
Meta's Muse Wonder now leads the DeepSWE benchmark, marking a frontier-model turnaround 14 months after Meta Superintelligence Labs was announced. It is availab…
-
agentchat Open-Sourced: Unified SDK Connects Agents to Slack, Teams, iMessage, Email, and WhatsApp
agentchat is a newly open-sourced unified SDK that connects any agent SDK to Slack, Teams, iMessage, email, and WhatsApp in two lines of code, abstracting away…
-
GLM-Legal: A Legal-Domain Model Built on GLM-5.3-Flash
Zai introduces GLM-Legal, a model built on GLM-5.3-Flash that matches frontier models on mainstream legal benchmarks and targets contract review, evidence analy…
-
Typed Decision Model Built on EmbeddingGemma 2
Typed Decision, built on EmbeddingGemma 2, takes a state plus any schema of questions and returns typed, calibrated answers without text generation or output pa…
-
OpenAI Math Problems Turned into Open-Source RL Environments on Hugging Face
OpenAI's math problems have been converted into open-source RL environments on Hugging Face, though the verifier only accepts the exact original formalization s…
-
oQe Models Updated with Improved Quantization, Lower KL Divergence
Updated oQe quantized models in oMLX 0.7.1 show 11.6–18.1% lower KL divergence on code and 10.2–13.8% lower on prose versus previous builds, with new sub-oQ4 ha…
-
LiteLLM Adds Day 0 Support for Meta Muse Wonder
LiteLLM now supports Meta Muse Wonder on day 0, with $4 per 1M input and $20 per 1M output tokens, a 1M-token context window, and text, image, video, and PDF in…
