Le Chonk AI vs GLM-5.3: documented capabilities

Versions and evidence

Read GLM-5.3 documentation, Z.ai API pricing for the other model. Listed rates are not normalized quotes: promotions, cache writes, long prompts and channel terms may change them.

Repair an invoice calculation without changing its public interface.

The function below handles an invoice. Fix rounding only at the final cent, reject negative quantities, preserve empty-input behavior, and add regression tests. Return a unified diff and test commands.

function invoice(items) {
  return items.reduce((sum, item) => sum + Math.round(item.price * item.qty * 100) / 100, 0);
}

Input A: [{price:0.335,qty:1},{price:0.335,qty:1}]
Input B: []
Input C: [{price:2,qty:-1}]

• A returns 0.67, not 0.68; B returns 0.
• C raises a documented validation error. Existing positive-quantity behavior stays intact.
• Execute tests in Node with no network. Give read/write/test tools only, at most 8 calls.

These checks are authored reference expectations, not either model’s output. For linked fixtures, download the operations PDF and chart.

Common evaluation protocol

1. Freeze the provider, exact model ID, date, reasoning setting and system prompt. Use a fresh conversation for every run.
2. Give both models the same task input and tool permissions. Record any provider-specific preprocessing and context truncation.
3. Use three attempts per task, a 60-second request timeout, and a fixed tool-step budget. Preserve failed/refused attempts as results.
4. Validate with the published checks. Record wall time, billed tokens, cache hits, tool fees and any human edits. Compare cost per accepted result.

Migration and billing caveats

GLM-5.3 documentation requires reasoning to remain enabled. Pin low/high/max rather than transferring a disabled-thinking option. Coding-plan quotas are not interchangeable with API token prices.

Cost per accepted task = all billed inference, tools and retries divided by accepted results. Keep output token usage and human corrections in the record; a lower per-token rate can still produce a more expensive job.

How to make a choice

For a text-only coding workflow, both belong in a task-level evaluation. If raw image input is required, the documented GLM-5.3 endpoint is not a direct substitute; preprocess the image or evaluate a separate vision model.

The next useful evidence is a preserved run trace from your own workload. Until it exists, this is a shortlist based on documentation. Browse task exercises.