OpenAI announced GPT-5 availability on its API platform (August 7, 2025) with claimed state-of-the-art performance on coding and agentic tasks: 74.9% on SWE-bench Verified, 96.7% on τ2-bench telecom tool-use benchmark, and reported 70% preference over o3 in frontend engineering internal tests. The model introduces new developer features including minimal reasoning effort mode, verbosity control, custom tool calling (plaintext instead of JSON), and parallel tool execution. Pricing: GPT-5 at 1.25/M input, 10/M output tokens; mini and nano variants at lower cost. The announcement emphasizes agentic reliability (chaining dozens of tool calls without failure) and reduced factual hallucination versus prior models. Financial institutions should track whether actual enterprise adoption and recognized revenue growth materialize, and whether agentic task reliability claims hold under production load conditions.
319 other tracked developments touching the same claims.