In my August 12, 2026 setup, starting a Codex task included a small purchasing decision: four model buttons, two usage buckets, and three token prices. The usage columns record that setup. I rechecked the API rates below on September 30; availability and preview allowances can change.
The model matrix
| Model | Context | Images | API input / output per 1M (Sept. 30) | Usage on my $200 Pro 20× plan | I would use it for | Main downside |
|---|---|---|---|---|---|---|
| GPT-5.6 Sol | 1,050,000 | Yes | $4 / $20 | Shared bucket | Long migrations, ambiguous debugging | Highest price |
| GPT-5.6 Terra | 1,050,000 | Yes | $2 / $12 | Shared bucket | Large repositories, everyday work | Middle choice still requires guessing |
| GPT-5.6 Luna | 1,050,000 | Yes | $0.20 / $1.20 | Shared bucket | Cheaper multimodal tasks | Uses the same shared allowance |
| GPT-5.3 Codex Spark | 128,000 | No | Not finalized | Separate preview bucket | Small edits, refactors, quick CSS work | Older model, 8.2× less context, no screenshots |
In my August setup, Spark had its own preview usage bucket. I could use it for small jobs without spending the shared allowance. The preview help page was blocked during this review, so I could not verify that the same allowance is available today. The choices below describe that setup.
The separate allowance is also a budget-design decision. A pooled cap encourages substitution; a separate bucket encourages experimentation. I saw the organizational version when my company combined several AI budgets into one pool.
The API rates above are standard text-token rates. The model pages specify higher rates for prompts over 272K input tokens, plus separate cache-write rates. Sol’s listed rate is promotional through at least November 21, 2026. A context-window size is not a promise that the entire window costs the base rate.
Historical estimate — . Not a current price quote.
1,000 requests × 1,000 input + 200 output tokens per request. Estimated standard text-token API cost in USD.
| Model | Input | Output | Total |
|---|---|---|---|
| GPT-5.6 Sol | $4.00 | $4.00 | $8.00 |
| GPT-5.6 Terra | $2.00 | $2.40 | $4.40 |
| GPT-5.6 Luna | $0.20 | $0.24 | $0.44 |
Historical rates, formula and assumptions
- GPT-5.6 Sol: $4 input / $20 output per million text tokens, as recorded September 30, 2026.
- GPT-5.6 Terra: $2 input / $12 output per million text tokens, as recorded September 30, 2026.
- GPT-5.6 Luna: $0.2 input / $1.2 output per million text tokens, as recorded September 30, 2026.
Cost = requests × (input tokens × input rate + output tokens × output rate) / 1,000,000. These are API costs, not deductions from the article's Pro-plan allowance. Spark is excluded because its API price was not finalized in this record.
Standard text rates only: no cache reads or writes, images, tools, batch discounts, taxes or other fees. Prompts above 272K input tokens use different rates and cannot be modeled here. Sol's recorded rate was promotional through at least November 21, 2026; this historical calculator does not verify its current availability. Equal token counts do not imply equal task quality or latency.
Exact input, output and total costs are in the table. With JavaScript, focus the chart and use arrow keys, or tap a model.
What I would choose
| Task | First choice | Why it fits | Usage consequence | What would make me switch |
|---|---|---|---|---|
| Rename, focused refactor, quick CSS edit | Codex Spark | Fast feedback for a small text-only change | Preserves the shared 20× bucket | The change spreads across the repository |
| Fix a UI from a screenshot or mockup | GPT-5.6 Luna | Image input at the lowest GPT-5.6 API price | Uses the shared bucket | The debugging requires more reasoning |
| Explore an unfamiliar large repository | GPT-5.6 Terra | 1.05M context with middle-tier pricing | Uses the shared bucket | The task is either trivial or unusually difficult |
| Plan a migration across many files | GPT-5.6 Sol | Maximum reasoning headroom and full context | Uses the shared bucket fastest | A smaller model proves sufficient |
| Continue after the shared limit gets tight | Codex Spark | Separate research-preview allowance | Leaves the shared bucket alone | I need images or more than 128k context |
Spark is fast and leaves the shared allowance alone, but it is text-only with roughly one-eighth of GPT-5.6’s context. A screenshot or a large repository sends me back to another button.
I now choose a model partly by which limit I am willing to hit. My Grok Build trial made that painfully concrete: its weekly ceiling changed whether I was willing to begin a task at all.
Was this useful?
Thanks for the feedback.