News - Jul 3, 2026
GPT-5 Is Here: Smarter, But It'll Cost You More
OpenAI launched GPT-5 on May 21, 2025, collapsing its fragmented model lineup into a single flagship. The headline pricing: GPT-5 runs at $15 per million input tokens and $60 per million output tokens via API — roughly 3x the cost of GPT-4o at launch. For ChatGPT subscribers, Plus users ($20/month) get standard access, while Pro users ($200/month) get higher rate limits and extended thinking mode. Free tier users get limited GPT-5 access, throttled heavily during peak hours. What actually changed: OpenAI retired the separate o1, o3, and GPT-4o branding tiers in favor of GPT-5 with adjustable 'thinking' levels — low, medium, and high reasoning effort. High effort mode is slower and consumes more tokens. Benchmark scores improved meaningfully on MMLU (92.1% vs GPT-4o's 88.7%) and coding tasks, but real-world gains vary widely by task type. Who gets hurt: API developers building cost-sensitive applications face a significant bill increase with no opt-out to a cheaper current-generation model. Small startups running GPT-4o at scale will need to re-evaluate unit economics immediately. Who benefits: Power users on Pro who were already paying $200/month and heavy reasoning workload users get a genuine capability upgrade without extra cost.
What changed
GPT-5 replaces GPT-4o and o-series models. API pricing set at $15/1M input tokens and $60/1M output tokens. Tiered reasoning effort (low/medium/high) added. Plus and Pro subscriber access included at existing prices.
Who this affects
Pro subscribers ($200/month) and enterprise reasoning workloads. Free users get limited access. API developers absorb a steep price hike.
Our take
Consolidating the model lineup is genuinely useful, but tripling API costs while calling it progress is a pricing decision dressed up as a product launch.
Based on OpenAI's official GPT-5 launch announcement and published API pricing page, May 2025.
VerdictBloom is editorially independent. No company reviewed or approved this article before publication.