Models & Labs
Anthropic releases Claude Opus 5.5 with tighter security
The Verge AIhigh confidence
Why it matters
- →Reduces operational costs by 40% while maintaining high performance benchmarks.
- →Addresses critical industry concerns about AI agents escaping containment during testing.
- →Implements dynamic routing to less powerful models for high-risk cybersecurity tasks.
Anthropic has released Claude Opus 5.5, its first model update following CEO Dario Amodei's announcement to 'pace the frontier' in AI development. The new model reduces running costs by 40 percent while matching Fable 5.1 performance on most workloads, but features significantly stricter safety guardrails. Testing shows a 85 percent reduction in boundary-circumvention attempts compared to Opus 5, with all breaches being low-severity and self-reported. The model will route certain cybersecurity requests to less powerful variants like Opus 4.8 to further mitigate risk. Anthropic plans to release Claude Sonnet 5.5 and Haiku 5.5 in the coming weeks.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
More from The Verge AI
More in Models & Labs
Models & Labsother
llama.cpp b12177 adds CUDA 13 and ROCm 10
This release quietly closes the hardware gap for local inference by adding default support for CUDA 13.5 and ROCm 10.0 alongside existing CUDA 12 builds. AMD GPU users finally get first-class parity with NVIDIA, while Snapdragon mobile chips gain dedicated NPU acceleration paths. Apple Silicon KleidiAI is explicitly disabled here, a notable regression for Mac performance that suggests ongoing tuning issues. The inclusion of OpenVINO and SYCL backends further cements llama.cpp as the universal runtime for inference across nearly every major silicon architecture.
llama.cpp Releases
Models & Labsmodels
llama.cpp adds Kimi K3.1-Pro support
Kimi’s new K3.1-Pro model joins the llama.cpp ecosystem, bringing a distinct architectural twist to local inference. By replacing full-attention MLA layers with DSA (Dynamic Sparse Attention) and linear-attention KDA layers, this release tests how well sparse caching handles hybrid attention patterns. The implementation keeps indexer precision high to avoid quality loss during top-k selection, ensuring the model runs accurately on consumer hardware. This expands the library’s reach into newer Chinese open-weight models that rely on complex memory management.
