ModelsGPT-5.2 Guide: xhigh Effort, GDPval and a 40% Price Rise
GPT-5.2 added an xhigh effort level, a 2025 knowledge cutoff and a 70.9% GDPval claim, and raised the price 40% over GPT-5.1. How to judge whether it earns that.
ReadNotes on models, agents, and what inference costs
Practical writing for developers building with large language models — how they work, how to pick one, and how to keep the bill predictable.
Picking between models — and testing them on your own work.
Models
ModelsGPT-5.2 Pro Guide: Paying $168 per Million Output Tokens
GPT-5.2 Pro is OpenAI's high-compute reasoning tier from December 2025. What it costs, why requests can take minutes, and when it still beats a cheaper model.
Read
ModelsGPT-5.2-Codex: What Changed From Codex-Max, and What It Cost
GPT-5.2-Codex added native compaction, better Windows and vision handling, and a small SWE-Bench Pro lead over GPT-5.2, at 40% more per token than Codex-Max.
Read
ModelsGPT-5.3-Codex Guide: The Terminal Specialist, Read Closely
GPT-5.3-Codex barely moved on SWE-Bench Pro but jumped on Terminal-Bench and OSWorld. What that pattern means, what it costs, and whether GPT-5.4 replaced it.
ReadModelsGPT-5.4 Guide: 1M Context, Computer Use and the 272K Cliff
GPT-5.4 merged OpenAI's Codex and GPT lines into one flagship with a 1.05M window. What it does well, where the long-context price doubles, and how to use it.
Read
ModelsGPT-5.4 Image 2: One Endpoint That Writes and Draws
GPT-5.4 Image 2 pairs GPT-5.4 with OpenAI's GPT Image 2 behind one endpoint. What the pairing is, how to call it, how to budget for it, and where it still fails.
Read
ModelsGPT-5.4 Mini Guide: Most of GPT-5.4 at Under a Third the Price
GPT-5.4 mini lands within a few points of GPT-5.4 on coding and computer use at $0.75 in and $4.50 out. Where the gap shows, and when mini should be your default.
ReadModelsGPT-5.4 Nano Guide: The Cheapest 5.4 Model and Its Hard Limit
GPT-5.4 nano codes almost as well as mini on SWE-Bench Pro but scores 39% on computer use. What it is built for, what it costs at volume, and when to step up.
ReadModelsGPT-5.4 Pro Guide: What Twelve Times the Price Actually Buys
GPT-5.4 Pro costs twelve times GPT-5.4 per token, runs only on the Responses API and has since been matched on price by GPT-5.5 Pro. When it still makes sense.
ReadModelsGPT-5.5 Guide: The Flagship That Now Costs More Than Its Successor
GPT-5.5 doubled GPT-5.4's price in April 2026, then GPT-5.6 Sol undercut it. What changed, what the defaults cost you, and when pinning it still makes sense.
Read
ModelsGPT-5.5 Pro Guide: When Six Times the Price Is the Cheap Option
GPT-5.5 Pro bills $30 in and $180 out per million tokens, defaults to high effort and can take minutes per answer. How to decide which calls deserve it.
Read
ModelsGPT-5.6 Luna Guide: Nano Pricing With a Six-Step Effort Dial
GPT-5.6 Luna is the cheapest tier of OpenAI's 5.6 family: nano-tier pricing, a 1M window and effort up to max. How to keep it cheap, and when GPT-6 Luna wins.
ReadShowing 109–120 of 446 articles