Guides, model notes and product updates

A ChatGPT subscription is a flat-rate seat for one human in the chat app; the API bills per token and has no subscription tier. Which is cheaper depends on your real invoice and your real per-request cost, not a remembered price.

A runnable FastAPI chatbot on one TokenLab API key, with SSE streaming, server-side memory, a model allowlist, error handling, dated per-1M-token prices and a smoke test.

A developer-focused guide to choosing the best AI video models API by workflow, async job handling, output format, and pricing.

The best AI models for coding in 2026 come down to routing: Claude Sonnet 5 for review, GPT-5.5 for premium agentic work, Gemini 3.5 Flash for long context, and DeepSeek V4 for cheap loops.

Picking an image generation API is a five-part decision: which operations the model supports, how it accepts reference images, which output controls it exposes, how it bills, and whether it fits a synchronous or task-based flow. This is a framework for making that decision and checking current prices yourself.

A corrected walkthrough of the TokenLab image edit endpoint: the right path, the two accepted image input shapes, multi-image support on gpt-image-2, and how async polling works when an edit is slow.