Seventeen focused tools, one evaluation workflow.

Fifty prompt fixtures, thirty raw outputs, and ten dated image-model comparisons.

A reproducible framework for latency, reliability, compatibility, and routing tests.

Compatibility checks and a practical migration path for OpenAI-style clients.

Compare response shape, route availability, and latency before changing providers.

Workload-based cost modeling for text, image, and video API usage.

Score providers against the reliability and compatibility requirements that matter to your product.

Measure latency, throughput, retries, and rate-limit behavior on endpoints you are authorized to test.

Estimate text, image, and video spend with caller-supplied current pricing.

OpenAI-compatible examples for routing text, image, and video model calls.

A shared API pattern for common LLM provider workflows.

Single and batch generation examples with retries and budget controls.

Async polling and batch prompt patterns for Gemini image workflows.

A resumable Python batch client with concurrency and retry controls.

Async video generation, polling, resume, and budget-limit examples.

成功率、P50/P95 延迟和 OpenAI 兼容性评测方法。

OpenAI、Claude 和 GPT API 中转兼容性检查清单。

按真实工作负载估算大模型与 AI API 成本。