According to SawyerMerritt, Grok 4.7 boosts self-checking, long-context handling, and pro-task benchmarks at same price and speed, per SpaceXAI.

According to DeepLearning.AI, Zai’s GLM-5.3-Flash hits 320B params, runs China-made hardware, and leads GDPval AA v2 at $0.09 per task.

According to OpenAI (@OpenAI), the company has launched GDPval, a new evaluation framework designed to measure artificial intelligence performance on real-world, economically valuable tasks. This new metric emphasizes grounding AI progress in concrete evidence rather than speculation, allowing businesses and developers to track how AI systems improve on practical, high-impact work. GDPval aims to quantify AI's effectiveness in domains that directly contribute to economic productivity, addressing a critical need for standardized benchmarks that reflect real-world business applications. By focusing on evidence-based evaluation, GDPval provides actionable insights for organizations considering AI adoption in operational workflows. (Source: OpenAI, https://openai.com/index/gdpval-v0)