Anthropic's Mythos 5.2 model exhibits incremental improvements in capabilities and performance, with notable advancements in literature review and computational analysis. However, the model still struggles with novelty and technical details, and its alignment risk remains "low" despite improvements. The model's cyber and malicious influence campaign capabilities are also under review, with some concerns raised about its robustness against prompt injection and attacks. AI summary
Firehose
Filtered to People, tagged “open-weight models” · clear filters
Browse: People · Companies · Papers · Podcasts · Hacker News · Deep dives
Browse by tag
OpenAI showcased new ways to work with Codex at the AI Engineer Code Summit, including voice as a user interface, cloud handoff, and improved dots. The company is working to address the awkward transition to cloud and make cloud functionality easier to use. AI summary
NaiveAI's Naive-N0.5-Flash model has been benchmarked against DeepSeek V4.1 Flash 1017 and GLM 5.3 Flash, and its ultrafast mode can peak at around 2,100 tokens per second and sustain nearly 1,000 during thinking-heavy tasks, with a median response time of several times shorter than GLM 5.3 Flash and DeepSeek V4.1 Flash 1017 at their high thinking settings. AI summary