US AISI says Kimi K3 trails top US frontier models on cyber evaluations
NandoDF · x · 2026-07-24
The US AI Safety Institute (AISI), together with NIST’s Center for AI Standards and Innovation, evaluated Kimi K3 on cyber capabilities.
- Their preliminary results say Kimi K3 performs below leading US frontier models on these cyber-focused evaluations.
- The post frames this as a real safety/measurement effort, not a product announcement.
- Because it is about a model’s security-related behavior under evaluation, it fits the policy/safety lane more than a general model-news lane.
More from Models
- Is DeepSeek's rumored K3 a scaled-down model, or something bigger? X users debate — teortaxesTex · 2026-09-11
- DeepSeek update keeps cache hits mid-conversation, cuts costs 36.6% — teortaxesTex · 2026-09-11
- AI Sextet offers 6 models free and unlimited for 14 days, including DeepSeek and Qwen — airesearch12 · 2026-09-11
- BullshitBench update: GPT-6-Astra beats all prior OpenAI models but still trails Anthropic — scaling01 · 2026-09-11
- Astra Scores 83% on GauntletBench, First Computer-Use Agent to Beat Human Baseline — ducha_aiki · 2026-09-11
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11