OSHAL any-bot · live

Cost & Token Race

Type a prompt, run it across every configured provider at once, watch cost / time / tokens / output land live, then open a graphical report. One runner, any provider & harness.

Educational & informational use only. Nothing here is private — treat every prompt and every result as public, shareable information. Don't enter confidential, personal, or sensitive data.

API logins

Add a key to enable a provider you don't have a CLI login for. Stored locally (gitignored), never committed.

All providers click any to configure it

Every provider in the registry. Green = ready (key set or CLI login); amber = needs a key you haven't added (e.g. xAI/Grok → XAI_API_KEY).

Configurations from the OSHAL provider registry (24 providers)

Pick a provider, choose a model, fill any required credentials, add it to the race. Direct-API providers run via their adapter; everything else (Bedrock, OpenRouter, xAI, Groq, …) routes through cline-cli.

Cost rate card USD / 1M tokens · cached = cached-input price

Estimates cost for providers that don't return dollars (codex / openai / gemini / cline). Claude uses real billed cost. Edit any cell, or load published list prices.

ModelInputCachedOutput
runs all providers concurrently · judges each answer

Cost per config (lower is better)

Latency

Total tokens

Answer quality (judged 1-10, higher is better)

Cost vs quality — top-left is best value (cheap + high quality)

Provider / ModelHarnessCostTok inTok outTimeQualityOutput