Scale AI releases SWE-Bench Pro V2
Scale AI released SWE-Bench Pro V2 on September 22, 2026. It is a refreshed public split of SWE-Bench Pro.
Related AI news
Scale AI open-sources AgentEnv for reinforcement learning environments
Scale AI has open-sourced AgentEnv, the foundational runtime used internally to power its reinforcement learning environments for training AI agents.
Scale AI brings NVIDIA Open Agent Safety Platform to Scale GenAI Portfolio
Scale AI is adding NVIDIA’s Open Agent Safety Platform technologies to the Scale GenAI Portfolio to build, train, deploy, run, and evaluate AI agents in controlled environments, with enforceable security and a record of their actions.
Scale AI and Google Cloud partner on a reference architecture for the Scale GenAI Portfolio
The joint blueprint brings the Scale GenAI Portfolio to Google Cloud and includes Gemini Enterprise integration. Scale AI says it is meant to give builders a proven foundation for putting AI to work faster.
DeepLearning.AI details DeepSeek cache reduction and Flash benchmark results in The Batch
In an issue of The Batch, DeepLearning.AI analyzed DeepSeek's architecture, reporting an 890-byte cache per token that is 437 times smaller than DeepSeek-V1. The analysis also notes a 25% compute increase per output token from 4K to 1M input tokens and highlights Flash outperforming V4-Pro on the AA Index.
Arena.ai evaluates Jev Router across more than 4,700 agentic sessions on Agent Arena
Arena.ai evaluated typesafeai's Jev Router across more than 4,700 agentic sessions on Agent Arena, finding that while it does not improve on the current Pareto frontier, it demonstrates strong steerability in response to user feedback.