DeepSeek has permanently reduced pricing for its V4 Pro model by 75%, bringing input token costs down to $0.003625 per million tokens from $0.0145. The move makes permanent a promotional discount that was set to expire May 31, 2026.
Anthropic's unreleased Claude Mythos Preview model has discovered more than 10,000 vulnerabilities across partner organizations in its first month of deployment through Project Glasswing. The company reports partners are finding bugs at 10x their previous rate, with Cloudflare discovering 2,000 bugs and Mozilla finding 271 Firefox vulnerabilities — 10x more than with previous Claude models.
Tencent released Hy-MT2, a family of multilingual translation models available in 1.8B, 7B, and 30B-A3B (MoE) sizes. All models support translation among 33 languages and follow translation instructions in multiple languages. The 1.8B model can be compressed to 440MB using 1.25-bit AngelSlim quantization.
NVIDIA has released Nemotron-Labs Diffusion, a family of diffusion language models at 3B, 8B, and 14B scales that generate multiple tokens in parallel rather than one at a time. The 8B model achieves 6.4× higher tokens per forward pass than autoregressive models in self-speculation mode while maintaining comparable accuracy.
Google announced two major features coming to its Gemini Mac app this summer: the Spark AI agent that can automate desktop workflows and access local files, and an enhanced voice control system. Spark will be available to Google AI Ultra subscribers ($100/month) and can integrate with Workspace apps and third-party services.
NVIDIA released Nemotron-Labs-Diffusion-14B, a 14-billion parameter language model that supports three decoding modes by switching attention patterns during inference. The model achieves 850 tokens per second on GB200 hardware at concurrency 1, representing a 3.3x speed-up over standard autoregressive decoding and outperforming Qwen3-8B-Eagle3 by 2.2x in self-speculation mode.
Google announced Gemini 3.5 Flash and Gemini Spark at I/O 2026. Gemini 3.5 Flash now powers Google's AI Mode search, while Spark is a cloud-based autonomous agent that can monitor credit card statements, track emails, and interact with third-party services like OpenTable and Instacart.
Tencent has open-sourced Hy-MT2, a family of multilingual translation models available in 1.8B, 7B, and 30B-A3B parameter sizes. The models support translation across 33 languages and include extreme quantization down to 1.25-bit, reducing the 1.8B model to 440MB storage while increasing inference speed by 1.5x.
Google has raised Gemini usage limits in its Antigravity coding tool by 3x twice within one week, responding to developers who hit new compute-based quotas within hours. The company also reset weekly quotas for all paid users twice, though limits remain lower than pre-restriction levels.
GitHub has released the source code for its Copilot for Eclipse plugin under the MIT license. The move makes Copilot's Eclipse integration the first open-source IDE plugin for the AI coding assistant, with code now publicly available on GitHub.
OpenAI has integrated ChatGPT into Microsoft PowerPoint, allowing users to generate and edit presentation slides using natural language prompts. The feature is available in public beta to both free tier users and ChatGPT Business subscribers.
Amazon Nova Act, AWS's browser-based AI agent service, now qualifies as HIPAA eligible, allowing healthcare organizations to deploy autonomous agents for workflows involving electronically protected health information. The service automates repetitive browser tasks including claims processing, referral coordination, and prior authorization.
Anthropic released version 0.104.0 of its Python SDK on May 21, 2026. The update adds support for a thinking-token-count beta feature that provides estimated token counts in thinking block deltas when streaming responses from reasoning models.
OpenAI released an update to Codex for Mac adding Appshots, a new feature that captures both screenshots and text from application windows using a Command-Command keyboard shortcut. The feature accesses content beyond what's visible on screen to provide fuller context to chat threads.
Google is expanding its 'Gemini built in' program to include speaker reference designs, allowing third-party manufacturers to build Gemini-powered smart speakers without lengthy development cycles. The program, which previously launched cameras through Walmart's Onn brand, now provides turnkey hardware solutions for both speakers and cameras.
Amazon Web Services released AgentCore Code Interpreter, a sandboxed Python environment that enables recursive language models to process documents of unlimited length by treating context as an external environment rather than loading it into the model's context window. The system orchestrates sub-LLM calls from within the sandbox, maintaining intermediate results as Python variables across a persistent session.
AWS has launched Amazon Bedrock AgentCore, a serverless runtime environment for deploying production AI agents. Turkish fulfillment company OPLOG demonstrated the platform's capabilities by building three business intelligence agents using Anthropic's Claude Sonnet, achieving a 35% reduction in sales cycles and 98% reduction in manual research time.
Perplexity has released a major update to its Comet AI browser for iOS, adding eight new features including one-tap phone number actions, a redesigned iPad sidebar, and Finance Deep Dive analysis that opens as browser tabs. The update also fixes persistent bugs with recently closed tabs and deleted conversation threads.
Replit today launched self-serve Enterprise, allowing organizations to purchase and configure an Enterprise account with SSO, SCIM, and RBAC in minutes without sales calls. The credit-based model provides unlimited seats, with all spending pooled across Replit Agent, Deployments, and Storage.
Alibaba has released Qwen3.7 Max, the flagship model in its Qwen3.7 series, featuring a 1 million token context window. The text-only model is designed for agent-centric workloads with strengths in coding, office productivity, and long-horizon autonomous execution, and includes explicit prompt caching support.