Author’s note: The 2K/4K tiers of gpt-image-2.6-all are upscaled tiers. The response may not echo size, and the actual pixel dimensions are occasionally off. This article explains why, and shows how to save the image, read its real width and height, and only then lay it out. A lot of people wire an image generation…
Author’s note: Once phones and browsers can synthesize speech locally, do you still need a cloud speech API? This article gives you a verdict across four dimensions: offline use, privacy, audio quality and long-form speech, and specifying a model with unified billing. More and more phone operating systems and browsers can read text aloud locally….
Author’s note: When you call a Large Language Model through a third-party API, answer quality alone can’t tell you which model is actually running. This post unpacks Anthropic’s open-source FlashProve: how it uses zkSNARKs to bind inference results to specific weights, what that costs, what it can and can’t prove, and how far it still…
Author’s note: Qwen4-Max is still in preview. It has no public benchmark scores, no stated context length, and no per-token pricing, so none of those three can be compared head-to-head with Qwen3.8-Max. The model you can actually call on a pay-as-you-go basis today is still qwen3.8-max. When the Qwen4-Max preview dropped, the first question from…
Author’s note: Grok Bot 1.0 is out of beta and now has team collaboration features, but you don’t get to choose the model. This article pulls together the official information and explains what to do if you need a fixed model ID and a fixed unit price. In its 2026-11-02 release notes, xAI introduced Grok…
TL;DR: I didn’t write any code. I just gave Codex requirements, approved its actions, and corrected its course, and it built me an Opus 5.5 one-sentence video workflow. The results are genuinely impressive, and the bill is genuinely not cheap. I did something pretty addictive. I had Codex recreate a video workflow: I type a…
Author’s note: This article compares three GPT-Image channel types (official, -all, and -vip) on billing, image sizes, and partial editing. It also reviews historical incidents from April to October 2026 and proposes a fallback plan built on limited retries, a circuit breaker, and a backup channel. If you’re integrating GPT-Image, you’ll quickly notice that the…
Author’s note: The Carbon introductory price ends on November 30. This article compares the Google blog footnote, a third-party price roundup, and the APIYI price list to explain whether thinking is billed as input or output, and how the output price change to $20 on December 1 affects your costs. Carbon’s introductory price has been…
Author’s note: this article explains the difference between Fable’s default of Medium in Claude Code and Fable 5.2’s default of high when you leave effort out of an API call. It also shows how to line up usage on both sides. A lot of developers start out writing code with Fable 5.2 in Claude Code…
Author’s note: A footnote from Google says that once the introductory period ends on 2026-11-30, the public-facing Gemini 4 Carbon moves to $4 input and $20 output. Argon doesn’t appear anywhere in the terms. This article breaks down the original wording and checks it against current pricing. If your project uses both gemini-4-carbon and gemini-4-argon,…
