Large language models systematically exclude religious perspectives when answering questions about grief, ethics, and family, according to new research from a multi-university consortium. Americans expected religion in AI responses 45-59% of the time depending on topic, but models mentioned it only 5-16% of the time.
StepFun has released Step-3.7-Flash, a 198B-parameter sparse Mixture-of-Experts vision-language model that activates approximately 11B parameters per token. The model supports a 256K context window, native image understanding via a 1.8B-parameter vision encoder, and offers three selectable reasoning levels.
NVIDIA has released Cosmos 3, a unified omni-model that combines world generation, physical reasoning, and action generation in a single architecture. Available in 8B (Nano) and 32B (Super) parameter versions on Hugging Face, Cosmos 3 uses a Mixture-of-Transformers architecture to process text, image, video, audio, and action modalities without switching between separate models.
MiniMax has released M3, a multimodal foundation model supporting text, image, and video inputs with a 1-million-token context window. The model costs $0.30 per million input tokens and $1.20 per million output tokens, available through OpenRouter.
OpenCode v1.15.13 introduces custom session metadata storage through its API and SDK. The release fixes a bug where Anthropic's Opus 4.7+ adaptive reasoning returned empty thinking blocks instead of summarized thinking.
Microsoft is ending GitHub Copilot's flat-rate subscription model in favor of token-based billing starting June 1. Some developers report monthly costs rising from approximately $29-50 to $750-3,000, while others claim the increases only affect inefficient "vibe-coders" who iterate excessively without clear direction.
Vercel released AI SDK version 4.0.0-canary.69 with breaking changes to xAI integration. The update deprecates the searchParameters option for xAI live search, replacing it with dedicated web_search and x_search agent tools, and adds image search capability through a new enableImageSearch parameter.
Vercel released version 5.0.0-canary.98 of its Google Vertex AI SDK, adding support for three new Google models: deep-research-max-preview-04-2026, deep-research-preview-04-2026, and gemini-embedding-2. The update enables developers to integrate Google's research-focused models and latest embedding model into applications using Vercel's AI SDK.
Vercel released version 4.0.0-canary.75 of its AI SDK Google package on May 30, adding support for three new Google models: gemini-embedding-2, deep-research-max-preview-04-2026, and deep-research-preview-04-2026. The update enables developers to integrate Google's latest embedding and deep research capabilities into applications built with the Vercel AI SDK.
OpenAI has expanded its Codex desktop app to Windows with Computer Use capabilities and remote control from ChatGPT mobile apps. The features, previously Mac-only, allow Codex to operate Windows desktop applications autonomously and enable iPhone, iPad, and Android users to initiate and monitor Codex tasks on Windows devices.
Google has launched Gemini Spark, a 24/7 AI agent for Google AI Ultra subscribers in the US. The service automates tasks across Google Workspace apps with remote browser control, supporting up to 15 concurrent tasks with compute-based usage limits.
Anthropic closed a $65 billion Series H at a $965 billion valuation, making it the most valuable AI startup globally and surpassing OpenAI's $852 billion March valuation. The company simultaneously released Claude Opus 4.8 and announced plans to bring its Mythos cyber-focused model to all customers within weeks.
StepFun has released Step-3.7-Flash, a 198B-parameter sparse Mixture-of-Experts vision-language model that activates 11B parameters per token and delivers up to 400 tokens per second. The model supports a 256K context window, three selectable reasoning levels, and is priced at $0.20 per million input tokens (cache miss) and $1.15 per million output tokens.
Liquid AI has released LFM2.5-8B-A1B, an 8-billion parameter hybrid model designed specifically for edge AI and on-device deployment. The model is available in multiple GGUF quantized formats ranging from 4-bit (4.84 GB) to 16-bit (16.9 GB), optimized for memory efficiency.
Google has modified Gemini's compute-based usage limits introduced at I/O 2026 after users reported depleting quotas too quickly. The company is now capping how much quota a single Gemini 3.1 Pro prompt can consume and making all 3.1 Flash-Lite prompts free.
StepFun has released Step 3.7 Flash, a 196B-parameter Mixture-of-Experts model that activates approximately 11B parameters per token. The multimodal model supports a 256K context window and introduces selectable reasoning levels (high/medium/low), priced at $0.20 per 1M input tokens and $1.15 per 1M output tokens.
Anthropic released Claude Opus 4.8 on May 28, 2026, replacing Opus 4.7 at unchanged pricing. The company claims the model's misalignment rates match those of Claude Mythos Preview, the experimental model deemed too dangerous for public release in April 2026. Opus 4.8 delivers faster thinking modes at one-third the cost of version 4.7.
Microsoft has rolled out a visual overhaul for Copilot in Microsoft 365, replacing the colorful interface with a predominantly black-and-white, text-forward design. The redesign, aimed at making the AI assistant feel "present but not imposing," includes a new adaptive prompt surface and consistent side panel placement across Word, PowerPoint, and Excel.
Microsoft is rolling out a redesigned Microsoft 365 Copilot that loads twice as fast, according to the company. The update introduces "progressive disclosure" — showing tools and controls contextually based on prompts rather than displaying all options at once.
Anthropic released Claude Opus 4.8 on May 28, 2026, with improved performance in agentic coding, computer use, and reasoning benchmarks. Pricing remains at $5 per million input tokens and $25 per million output tokens, while the model's fast mode is now three times cheaper than previous versions.