# Cost control

> Where the tokens actually go in an agentic workload, and the four levers that move the bill.

_Source: https://aiprimetech.io/docs/guides/cost-control/ · Home > Docs > Working with the API_

If you have run a coding agent for a full working day you have seen the bill. Agentic loops are token-hungry in a way ordinary chat never is: every file read, every tool result and every diff lands in context and is resent on the next turn.


## Output tokens are the expensive ones

On every frontier model, output costs 4–5× input. A model that thinks out loud in its final answer burns money faster than one that answers tightly. Cap `max_tokens` deliberately and prefer concise models for routine work.

| Model | $/M in | $/M out | Ratio |
|---|---|---|---|
| `claude-opus-4-8` | $3 | $15 | 5× |
| `claude-sonnet-4-6` | $0.90 | $4.50 | 5× |
| `claude-haiku-4-5` | $0.40 | $2 | 5× |
| `gemini-3-flash` | $0.20 | $1.20 | 6× |


## The four levers

1. **Route by task.** Classification, extraction and routing do not need Opus — Haiku does them at roughly a seventh of the price. Reserve the flagship for reasoning-heavy work.
2. **Prompt caching.** Stable system prompts and tool definitions are re-read from cache at a fraction of the input price. For agents this is the single biggest lever — see [Prompt caching](/docs/guides/prompt-caching/).
3. **Trim context each turn.** Do not resend an entire transcript; keep recent turns plus a running summary. Unbounded context growth is what turns a $2 session into a $20 one.
4. **Cap the loop.** Set a hard step limit on autonomous runs so a stuck agent cannot spin forever.


## Estimating before you run

```python
IN_RATE  = {"claude-opus-4-8": 3.0,  "claude-sonnet-4-6": 0.9, "claude-haiku-4-5": 0.4}
OUT_RATE = {"claude-opus-4-8": 15.0, "claude-sonnet-4-6": 4.5, "claude-haiku-4-5": 2.0}

def cost(model, tok_in, tok_out):
    return tok_in / 1e6 * IN_RATE[model] + tok_out / 1e6 * OUT_RATE[model]

print(cost("claude-opus-4-8",   20_000, 2_000))   # 0.09
print(cost("claude-sonnet-4-6", 20_000, 2_000))   # 0.027
print(cost("claude-haiku-4-5",  20_000, 2_000))   # 0.012
```

The same 20K-in / 2K-out task costs 7.5× more on Opus than on Haiku. If the task is classification, that multiple buys nothing.


## Measuring after

Every response carries a `usage` object. Log it against a request id and a task type — without that, cost optimisation is guesswork. The gateway dashboard shows per-model usage, but only your own logs know *why* a request was made.

> The most common expensive mistake is not the model choice — it is resending a growing transcript on every turn of a long agent session. Fix context growth first, then tune models.

- [Prompt caching](https://aiprimetech.io/docs/guides/prompt-caching/) — The biggest single lever
- [Context management](https://aiprimetech.io/docs/guides/context-management/) — Keeping history bounded
- [Models](https://aiprimetech.io/docs/getting-started/models/) — Rates for every model


---
_ClaudeAPIKey.dev is an independently operated, Anthropic-compatible API gateway. Not affiliated with Anthropic._