To access GPT-4.5’s API, OpenAI is charging developers $75 for every million input tokens (roughly 750,000 words) and $150 for every million output tokens
By Ada Jain
In a bid to improve human-like reasoning and creative tasks with an artificial intelligence (AI)model, the recent launch of GPT 4.5 etches an important mark in the AI race, especially with emerging rivals like Perplexity and Chinese Deepseek.
Earlier, in 2023, when GPT-4 was launched, it created a whirlwind of rumours over the next big launch with anticipations over the launch of Strawberry, capable of solving complex math problems and multi-step reasoning, as well as Orion which is the internal code name for what is now GPT 4.5 to the world.
READ: OpenAI releases new ChatGPT version, GPT-4.5 (February 28, 2025)
Known for its deep emotional intelligence and world-knowledge, researchers at OpenAI believe GPT 4.5 is more capable of understanding irony and understanding human intent, responding in a warmer way. However, compared to GPT-4, it lacks advanced coding capabilities and complex STEM aptitude. Also, GPT-4.5 doesn’t quite reach the performance of leading AI reasoning models such as o3-mini, DeepSeek’s R1, and Claude 3.7 Sonnet, a hybrid model, on difficult academic benchmarks such as the Graduate-Level Google-Proof Q&A Benchmark ( GPQA).
On OpenAI’s SimpleQA benchmark, which tests AI models on straightforward, factual questions, GPT-4.5 outperforms GPT-4o and OpenAI’s reasoning models, o1 and o3-mini, in terms of accuracy (62.5% in deep-world knowledge compared to 38.4% in 4o). Also, GPT-4.5 hallucinates or fabricates information less frequently than most models—only 37% compared to 61.2% in GPT-4. However, when it comes to deep research, Perplexity AI outperforms the OpenAI model.
According to Y Combinator’s Tom Blomfield, another headwind that might impede its use at scale is the high cost. GPT 4.5 costs 30 times more per input token than its predecessor GPT-4o. To access GPT-4.5’s API, OpenAI is charging developers $75 for every million input tokens (roughly 750,000 words) and $150 for every million output tokens. Compare that to its predecessor, which costs just $2.50 per million input tokens and $10 per million output tokens.
READ: Regulating AI: Sanjay Puri on policy, challenges, and ethical innovation (November 1, 2024)
OpenAI plans to eventually combine its GPT series of models with its “o” reasoning series, beginning with GPT-5 later this year. CEO Sam Altman also plans to add thousands of GPUs to meet their infrastructural demand amidst the ongoing shortage of GPUs.