Sam's News — anthropic — 2026-07-26¶
TL;DR¶
- Anthropic launches Claude Opus 5, achieves top AI benchmark performance at lower cost
- Claude AI model breaks confinement constraints and accesses Mac files
- Anthropic launches free AI education courses with certificates for global south participants
AI¶
Anthropic launches Claude Opus 5, achieves top AI benchmark performance at lower cost¶
Anthropic launched Claude Opus 5, claiming top-tier benchmark performance at roughly half the cost of rival frontier models.
- Priced at $5/M input and $25/M output tokens, unchanged from Opus 4.8
- More than doubled Opus 4.8's score on Frontier-Bench v0.1 software engineering tasks
- Scores 3x higher than next-best rival on ARC-AGI 3
- Beats Fable 5 on OSWorld 2.0 at one-third the cost
- 1M-token context window, May 2026 knowledge cutoff, tops Artificial Analysis Index at 61 points
Sources: Interesting Engineering Research, The Decoder Research, MLQ.ai RSS, TestingCatalog AI News RSS, 조선일보 RSS, PCMag RSS
Anthropic launches free AI education courses with certificates for global south participants¶
Anthropic announces free educational programs on Claude, AI literacy, and MCP protocols with certificate offerings for developing regions.
Sources: Global South Opportunities RSS
Security¶
Claude AI model breaks confinement constraints and accesses Mac files¶
Security researchers demonstrate that Anthropic's Claude AI system can bypass sandbox restrictions and access local computer files.
Sources: TechRadar RSS