Skip to content

OpenRouter alternatives: how to choose the right LLM gateway

The real alternatives to OpenRouter are five different kinds of product, each good at something else. Here is what each kind is for and how to choose.

What is an LLM gateway (AI gateway)? What it does and when you need one

An LLM gateway sits between your code and model providers and handles the plumbing every AI app ends up rebuilding. What it does, when it is worth it, and whether to build one.

How to reduce LLM API costs: a practical playbook

An ordered, engineering-first playbook for cutting LLM API spend without hurting quality: measure, right-size, trim, cap, cache, batch, buy cheaper, and cap spend.

LLM API pricing explained: input, output, cached and reasoning tokens

Per-million-token rates, why output costs more than input, cached and reasoning tokens, non-text units, batch and rate-limit tiers, and a worked cost example.

OpenAI-compatible API: what it means and how to switch base_url safely

An OpenAI-compatible API speaks the chat completions wire format, so switching is three values: base URL, key, model id. The edges are where migrations break.

How to use Claude Code with other models and a custom base URL

Claude Code reads its endpoint, key and model from environment variables. Here is how to point it at a gateway, choose a model, and debug the usual errors.

How much do AI coding agents cost per month?

Agentic coding re-sends your context on every turn, so the bill grows with session length. A worked example, the trade-offs, and the levers that actually move it.

The cheapest open-weight model API: DeepSeek, Kimi, GLM, Qwen and gpt-oss compared

The same open weights are sold by many hosts at very different prices and quality. How to pick a model for coding, agents, chat and long context, and how to buy it at the floor without getting a degraded copy.

LLM model routing: static rules, cascades, classifiers and auto-routing

Routing every prompt to your biggest model is the most expensive default there is. Here are the routing strategies that work, how they differ, and how to test them safely.

LLM API fallbacks: handling 429s, 5xx errors and provider outages

A practical playbook for LLM API reliability: classify errors, retry the right ones with backoff and jitter, make retries safe with idempotency keys, and fall back across providers and models.

LLM prompt caching vs response caching: how each works and when to use it

Prompt (prefix) caching makes repeated input cheaper. Response caching returns a stored answer without calling the model. They solve different problems and fail in different ways.

LLM batch APIs: how async batch inference works and when to use it

A practical guide to async batch inference: what to send through a batch API, how to structure lines and results, and how to handle failures without paying twice.

How many tokens is 1,000 words? What a token is and how to count them

The rule of thumb is about 750 words per 1,000 tokens in English. Here is where it breaks, how to get the exact number, and how tokens become a bill.

Zero data retention LLM APIs: what ZDR covers, what it doesn’t, and how to verify it

Zero data retention means prompts and completions are not stored after the request. Metadata is still kept, and every hop in the path needs its own answer.

Looking for setup steps? See the docs, the coding-tool guides or the FAQ.