can you find performance benchamrkts for models especially openai 5.4?

OpenAI's GPT-5.4 demonstrates strong performance on several benchmarks, particularly in professional and knowledge-work tasks. On the GDPval benchmark, which evaluates AI across 44 professions like accounting, sales, and engineering, GPT-5.4 achieved an 83% score, marking a 20% improvement over GPT-5.2 at similar costs [1][2][5]. It also set records on OSWorld-Verified (75.0%, surpassing human performance of 72.4% for desktop navigation via screenshots and actions) and WebArena Verified, showcasing gains in reasoning and computer use [2][7]. Additionally, GPT-5.4 generates output at 77.6 tokens per second via OpenAI's API, exceeding the median of 57.6 t/s for similar reasoning models [10]. For other models, China's GLM-5 (a 744-754 billion-parameter open-source model) reaches 80-90% of frontier-model performance and outperforms competitors like Gemini 3 Pro and Grok 4 on unspecified benchmarks, while offering lower costs at $1 per million input tokens [6][12]. Sources lack detailed cross-model comparisons beyond these highlights.