#benchmarks

How Agent Anomaly Detection scores on benchmarks we didn't write

Agent Anomaly Detection (AAD) provides asynchronous, low-latency monitoring for AI agents by analyzing OpenTelemetry traces and logs to flag goal hijacks and tool misuse. Tested across five public security benchmarks, AAD successfully caught 87% of harmful instructions carried out by agents without adding runtime latency. Developers can integrate AAD alongside semantic governance and strict tool limits as part of a defense-in-depth strategy to secure agentic applications.

Google Developers Blog · EN

AMD EPYC 9686F Linux Benchmarks: 96 Zen 6 Cores At Up To 5.3GHz

In addition to the AMD EPYC 9996 256-core CPU benchmarks out today in that dual socket configuration for a combined 512 cores / 1024 threads flagship Zen 6C server configuration, also tested recently at Phoronix was the EPYC 9686F part. The AMD EPYC 9686F is their new high-frequency Zen 6 part that features 96 cores / 192 threads at a boost clock of 5.3GHz! When the AMD EPYC 9686F they originally disclosed a 5.0GHz boost clock but with how well their production silicon panned out and/or marketing reasons, it's actually now confirmed to be 5.3GHz in production.

Phoronix · EN

AMD EPYC 9996 Benchmarks Show Unparalleled Performance With 256 Zen 6C Cores

Back in July AMD announced the EPYC 9006 "Venice" series at their Advancing AI event. Today they are announcing EPYC 9006 series Socket SP7 processor availability and with systems beginning to ship today from their OEM/ODM partners. With that we are also now allowed to begin sharing AMD EPYC 9006 performance benchmark figures. Back in July I expected EPYC 9006 to be a "grand slam" for performance... After hands-on with the 6th Gen EPYC processors, it's a grand slam alright -- out of this world.

Phoronix · EN

Show HN: Turn JMH Benchmarks Results into Interactive Charts

Visually explore your JMH Benchmarks. Contribute to mlangc/jmh-visualizer development by creating an account on GitHub.

Show HN · EN

Show HN: Slopgrid – create your own AI model benchmarks without code

Create your own AI model benchmark: generate multiple LLM outputs at once and compare them side-by-side

Show HN · EN