Apr 25
Loading video player...
GPT-5.5 Is Here: Benchmarks, Codex Agents, Context Window & Pricing Explained The video reviews OpenAI’s newly released GPT-5.5, now rolling out to ChatGPT and Codex, positioned as a “new class of intelligence for real work” with agent capabilities to understand complex goals, use tools, and carry tasks to completion, including computer and browser use. The host highlights that OpenAI led the announcement with Codex and notes GPT-5.5’s strengths in writing/debugging code as well as researching online, analyzing data, and creating documents and spreadsheets. Benchmarks discussed include GDP Val (84.9% across 44 professions) and OSWorld (78.7%, above the 72.4% human baseline), plus mixed coding results where Opus 4.7 leads SWE Bench Pro. GPT-5.5 matches GPT-5.4 latency, uses fewer tokens for Codex tasks, supports up to a 400,000 context window, and is priced higher in the API ($5/$30 per million input/output tokens; Pro $30/$180). 00:00 GPT 5.5 Arrives 00:29 Codex First Focus 00:57 Speed and Token Efficiency 01:28 Real World Benchmarks 01:58 What Work Remains 03:10 Demos and 3D Games 03:55 Spreadsheets and Browser Use 04:50 Eval Results and Rivals 05:32 Tiers Context and Pricing 06:33 Cost Justification Wrap Up 06:57 Final Thanks and Outro