Try CodeRouter's routing in your browser
Send coding requests from this page and see which model answers each one, and why: the task phase, how hard the task looks, what the judgment model was asked, and the four tiers with their list prices. They are real requests, through the same router and at the same list prices as requests from Claude Code, Codex or Cursor.
No credit card. On the free trial the playground takes up to 20 requests a day, billed at model list prices from the trial's $5 credit.
Preview a route
Paste a coding task to see which model smart routing would start it on, and why. No account needed; no coding model runs and nothing is billed.
Your text goes to CodeRouter's routing judge (the Jev model, run by TypeSafe AI) to classify it. It isn't stored, and no coding model runs. The bot check is Cloudflare Turnstile. Privacy
This retry loop sometimes runs forever when the API returns 429. Find the bug and fix it.
async function fetchWithRetry(url, opts) {
let attempt = 0;
while (attempt < MAX_RETRIES) {
const res = await fetch(url, opts);
if (res.status === 429) {
attempt = 0;
await sleep(nextDelay(attempt));
continue;
}
if (!res.ok) {
attempt++;
continue;
}
return res;
}
throw new Error("too many retries");
}The 429 branch sets attempt = 0, so the loop never reaches MAX_RETRIES. Count every retry and wait before the next one:
if (res.status === 429) {
attempt++;
await sleep(nextDelay(attempt));
continue;
}Presets to start from
- Small edit
- Find a bug
- Plan a refactor
- Write tests
Each needs a different amount of work, so the route can differ. Up to 20 turns per chat; answers up to 2,048 tokens.