July first-week open AI toolchain cluster: Anthropic two-angle + HF three-cloud + transformers 8 models + ExecuTorch export

“'The first week of July wasn't a model week — it was a toolchain week. Five releases, one stack: open-weight models from Hub to cloud to agent framework to export target.'”

Usable claims

• Anthropic's July 6, 2026 features post documents that Claude Code was built first as an internal CLI tool at Anthropic, used by engineers in the terminal before being released more broadly; the engineering team deliberately chose terminal-native ergonomics over chat interfaces, with minimal surface area, transparent tool use, and tight integration with existing developer environments (file editing, test running, repository navigation) rather than a separate IDE.
• Hugging Face's July 7, 2026 blog post announces a deep-link integration with Amazon SageMaker AI: supported HF models expose 'Customize on SageMaker AI' and 'Deploy on SageMaker AI' buttons that open the corresponding SageMaker Studio workflow with the model context preserved; new Studio environments are auto-provisioned with the AmazonSageMakerModelCustomizationCoreAccess managed policy covering SFT, DPO, RLVR, and RLAIF fine-tuning approaches plus deployment to SageMaker AI or Bedrock endpoints; the instance selection UI surfaces GPU quota availability (G5, G6 families) inline.
• Hugging Face's July 7, 2026 blog post documents that Microsoft Build 2026 announced Foundry Managed Compute with a curated catalog of Hugging Face open-weight models available directly in the Foundry Model Catalog — refreshed weekly, deployable in one click, spanning text/vision/audio/multimodal (LLMs, VLMs, ASR, embeddings, segmentation, image generation) — screened for license compliance and security and shipping only in SafeTensors format. Microsoft pre-stages weights in Azure storage, builds and CVE-scans runtime images (vLLM, SGLang, TensorRT-LLM, NIM, TEI, llama.cpp, hf-serve), and validates each model + runtime + accelerator combination before publication. Available in preview on NVIDIA A100, H100, and AMD MI300X accelerators in Global and Data Zone scopes.
• Hugging Face's July 7, 2026 blog post documents that Hugging Face Storage is now a first-class SkyPilot backend via a new `store: hf` option and `hf://` URLs: Buckets (read-write) or any model/dataset/Space repo (read-only) mount into a SkyPilot task using the `hf-mount` FUSE driver with MOUNT and COPY modes and lazy reads; data can be read onto GPUs across 20+ clouds, Kubernetes, Slurm, and on-prem environments without per-cloud copies or transfer taxes (HF charges no egress or CDN fees). A Qwen3.5-4B SFT benchmark loaded free in about 30 seconds at up to 500 MB/s, with checkpoints streaming to Buckets at 112-168 MB/s across AWS, GCP, and Lambda. Buckets are Xet-backed with content-defined chunking for incremental checkpoint deduplication.
• huggingface/transformers v5.13.0 (released 2026-07-03) adds 8 new models: Kimi K2.5/2.6/2.7 (multimodal agentic architecture from Moonshot for long-horizon coding, front-end design, and swarm-based task orchestration across Rust, Go, and Python); Xiaomi MiMo-V2-Flash (27T-token MoE with native 32k context and extendable 256K window); NVIDIA Nemotron 3.5 ASR (multilingual) and Nemotron ASR Streaming (English, cache-aware FastConformer-RNNT with configurable 80/160/560/1120 ms chunk sizes); Qwen3 ASR (Whisper-style encoder + Qwen3 decoder with forced-aligner head); Zyphra ZAYA1 (760M active / 8.4B total MoE, Compressed Convolutional Attention); Google DeepMind VideoPrism (general-purpose video encoder pretrained on 36M video-caption pairs and 582M noisy clips); NVIDIA RADIO (distills CLIP, DINOv2, and SAM into a single variable-resolution ViT); OpenBMB MiniCPM3-4B (MLA from DeepSeek-V2 + SwiGLU).
• huggingface/transformers v5.13.0 (released 2026-07-03) introduces the HfExporters subsystem: a unified base class with DynamoExporter, OnnxExporter, and ExecutorchExporter subclasses giving a single API for PyTorch/ONNX/ExecuTorch export with automatic prefill/decode splitting for generative models; the release standardizes layer declarations, mask/cache construction, and hybrid-attention handling to make models cleanly ONNX-, torch.export-, and ExecuTorch-exportable.