v1.0.1

Changes

• Added GitHub badges (stars, release, license, last commit) to README
• Improved documentation discoverability

v1.0.0

Initial Release

Features

• YAML-driven benchmark configuration
• Latency benchmarks (TTFT, P50, P99) with configurable concurrency
• Throughput testing under sustained load
• Cost analysis with 2026 pricing for 20+ models (OpenAI, Anthropic, Google, DeepSeek, Mistral, Meta, Xiaomi, Qwen)
• Quality scoring with automated judge model evaluation
• HTML, JSON, and Markdown report generation
• CI/CD integration with threshold gates (ci-check command)
• Model discovery from any OpenAI-compatible endpoint

Quick Start

pip install llm-api-bench
llm-api-bench run --config bench.yaml

Works with XiDao Gateway — a single endpoint for 50+ models.