Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Benchmarks token generation speed across multiple LLM API providers by measuring TTFT, tokens-per-second throughput, and total generation time. Configuration is read from ~/.openclaw/openclaw.json, with support for anthropic-messages, openai-completions, and openai-responses API formats.
Measures TTFT separately from total generation time, giving a more accurate picture of perceived latency than simple end-to-end timing scripts.
When to use it
- Comparing TTFT between OpenAI and Anthropic for a latency-sensitive app
- Verifying a new API key and endpoint are reachable before deploying
- Measuring throughput differences between two versions of the same model
- Checking whether a self-hosted LLM endpoint matches a hosted provider's speed
- Running repeated benchmark runs to catch API performance regressions
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Compare token generation speed across all configured providers
- 1Reads provider definitions from ~/.openclaw/openclaw.json
- 2Runs preflight connectivity check on all targets
- 3Sends short, medium, and long prompts to each provider
- 4Records TTFT, TPS, and total time per run
- 5Outputs results as a formatted comparison table
Table showing TTFT, tokens/sec, and total time for each configured provider and model
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.