⚙️ API & Deprecations
Anthropic launches Claude Sonnet 5 with introductory pricing of $2/$10 per MTok through August 31, 2026, after which standard pricing of $3/$15 applies. The model supports a 1M token context window and 128k max output tokens. (Anthropic API — June 30, 2026)
Anthropic has restored access to Claude Fable 5 and Claude Mythos 5 after a previous incident. (Anthropic API — July 1, 2026)
🚀 Models & Releases
Claude Fable 5 and Claude Mythos 5 were previously launched on June 9, 2026, as Anthropic's most capable widely released models, also featuring a 1M token context window and 128k max output tokens. (Anthropic API — June 9, 2026)
Poolside released Laguna XS 2.1, a new 33B-A3B category coding agent model, available both as a free and a standard version. (Poolside: Laguna XS 2.1 (free), Poolside: Laguna XS 2.1)
New models have been published on Hugging Face, including ByteDance-Seed/PAR and Microsoft's HARC-aligned models based on Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct. (ByteDance-Seed/PAR, microsoft/HARC-Qwen2.5-7B-Instruct, microsoft/HARC-Llama-3.1-8B-Instruct)
🛠️ Tools & Products
llama.cpp added a new public C API function llama_model_ftype_name() to expose a model's quantization type name (e.g., "Q8_0"). ([ggml-org/llama.cpp] b9860](https://github.com/ggml-org/llama.cpp/releases/tag/b9860))
Goose (v1.41.0) added support for iFlytek Spark, Astron MaaS, and Fireworks AI providers, along with OpenRouter request parameters and new desktop locales. ([block/goose] v1.41.0](https://github.com/aaif-goose/goose/releases/tag/v1.41.0))
Pydantic-AI released versions 2.3.0 and 2.4.0, adding a native Zhipu AI provider, a GEval evaluator, and security controls for file uploads. ([pydantic/pydantic-ai] v2.3.0 (2026-07-01)](https://github.com/pydantic/pydantic-ai/releases/tag/v2.3.0), [pydantic/pydantic-ai] v2.4.0 (2026-07-02)](https://github.com/pydantic/pydantic-ai/releases/tag/v2.4.0))
The Anthropic Python SDK was updated to v0.115.1, removing some nonfunctional types. ([anthropics/anthropic-sdk-python] v0.115.1](https://github.com/anthropics/anthropic-sdk-python/releases/tag/v0.115.1))
OpenAI's Codex released version 0.143.0-alpha.35. ([openai/codex] 0.143.0-alpha.35](https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.35))
🔬 Research
New research highlights security risks in persistent AI coding agents, where attacks can be distributed across pull requests and timed for deployment. (Distributed Attacks in Persistent-State AI Control)
A new testbed, LACUNA, was introduced for evaluating localization precision in LLM unlearning methods, which aim to remove sensitive data like PII. (LACUNA: A Testbed for Evaluating Localization Precision for LLM Unlearning)
A paper proposes "Program-as-Weights," a paradigm for implementing fuzzy functions locally to avoid reliance on costly LLM APIs for tasks like log alerting or JSON repair. (Program-as-Weights: A Programming Paradigm for Fuzzy Functions)
🏢 Industry & Business
Mark Zuckerberg reportedly told Meta staff that AI agent development has not progressed as quickly as he had hoped. (Mark Zuckerberg tells staff that AI agents haven’t progressed as quickly as he’d hoped)
An article criticizes the pervasive "AI hype," citing its mention in a sandwich chain's IPO documents as an example of overreach. (Jersey Mike’s IPO illustrates how bad the AI hype has become)
Google hosted an AI summit with New York City educators and industry leaders to discuss the future of AI in classrooms. (New York City educators and industry leaders gathered at Google’s offices to shape the future of AI in classrooms.)
Vinton Cerf, a key creator of the internet's foundational protocols, is retiring from his role as Google's chief internet evangelist. (The ‘Father of the Internet’ is finally retiring)
⚖️ Regulation & Society
Peter Thiel controversially labeled Pope Leo XIV a 'Chinese Communist Agent' over the Pope's stance on AI regulation. (Peter Thiel Brands Pope Leo XIV 'Chinese Communist Agent' Over His AI Regulation Stance)
Arizona student leaders are preparing an AI policy that could be adopted by schools in the state. (Arizona student leaders prepare to pass AI policy that could be utilized in Arizona schools)
An article outlines six key AI policy issues for growing businesses to consider before scaling. (Your First AI Policy: Six Burning Issues for Every Growing Business to Consider Before You Scale)
💡 Key Takeaway
Anthropic's launch of Claude Sonnet 5 at a discounted introductory rate provides a cost-effective, high-context window option for the next two months, making it a prime candidate for testing in production workloads before the price increases.