OpenAI: GPT-5.4 Nano

GPT-5.4 Nano is a paid OpenAI model that accepts text, images, and files as input, supports tool use and reasoning, and offers a 400,000-token context window with up to 128,000 tokens of output. Structured output support is unconfirmed. The combination of long context and multimodal input makes it workable for document-heavy or mixed-media workflows where reasoning over large inputs matters. On price, $0.20 per million input tokens and $1.25 per million output tokens positions it toward the budget end of the reasoning-capable tier, which is its clearest practical argument. Benchmark coverage is limited to three benchmarks, so the blended score of 68.0 should be treated as a rough signal rather than a settled verdict; coding scores (92.5) look strong, but agentic performance (45.4) is notably lower. Teams running coding-heavy workloads on a tight budget have the most reason to shortlist it, while those prioritizing agentic tasks may want to compare alternatives before committing.