GPT-5.6 Terra vs Claude Sonnet 5 pricing.
The two default choices for production work. They match exactly on input price and diverge on everything else.
- Input gap
- same
- Output gap
- 1.2×
- Cached gap
- same
- Rates verified
- 2026-10-02
Current-model note: Both models are current. The calculator separates exact local OpenAI counts from calibrated Claude estimates.
| Current model | Input / 1M | Cached / 1M | Cache write / 1M | Output / 1M | Context |
|---|---|---|---|---|---|
| GPT-5.6 TerraOfficial pricing ↗ | $2.00 | $0.20 | $2.50 | $12.00 | 1,050,000 |
| Claude Sonnet 5Official pricing ↗ | $2.00 | $0.20 | $2.50 | $10.00 | 1,000,000 |
Rates verified 2026-10-02. The browser may refresh them from the live catalogue. Provider pricing and tiers remain authoritative.
GPT-5.6 Terra: exact local BPE tokenization
Claude Sonnet 5: Provider-Calibrated UTF-8 Projection
PDF, DOCX, code, text, and data files are extracted locally and applied to both models.
GPT-5.6 Terra: documented formula
Claude Sonnet 5: documented formula
Prompt text, document contents, and image pixels stay in the browser.
Text, code, PDF, DOCX, images, audio, or video · 12 MB per file
Your input, measured across providers.
Choose a provider card to inspect every available model.
Workload & Scaling Planner +
What five real workloads actually cost
Headline per-token prices rarely decide a bill. Request shape does. These figures are calculated from the verified rates for GPT-5.6 Terra and Claude Sonnet 5, including any long-context tier that applies once a request crosses its threshold.
| Workload | Input | Output | GPT-5.6 Terra | Claude Sonnet 5 | Difference |
|---|---|---|---|---|---|
| Short chat turn A typical assistant exchange. | 1,000 | 500 | $0.0080 | $0.0070 | 1.14x cheaper on Claude Sonnet 5 |
| RAG answer Five retrieved chunks plus a question. | 12,000 | 800 | $0.034 | $0.032 | 1.05x cheaper on Claude Sonnet 5 |
| Code review A medium pull request with surrounding files. | 60,000 | 2,000 | $0.144 | $0.140 | 1.03x cheaper on Claude Sonnet 5 |
| Whole-document analysis A long report or contract read in one call. | 300,000 | 4,000 | $1.27 (tier) | $0.640 | 1.99x cheaper on Claude Sonnet 5 |
| Full-context load Filling most of a one-million-token window. | 900,000 | 4,000 | $3.67 (tier) | $1.84 | 2.00x cheaper on Claude Sonnet 5 |
"(tier)" marks a request that crossed a long-context threshold, so it is billed above the headline rate. Output length is the assumption most worth changing for your own case: paste a real prompt into the calculator above to replace these with your numbers.
The cached-input rate is the number most comparisons miss
Agents, chat threads and RAG pipelines resend the same system prompt and context on every call. Those repeated tokens bill at the cached rate, not the headline rate, so a model can be cheaper on paper and more expensive in production.
| Model | Fresh input / 1M | Cached input / 1M | Discount | Break-even |
|---|---|---|---|---|
| GPT-5.6 Terra | $2.00 | $0.200 | 90% | 1 read |
| Claude Sonnet 5 | $2.00 | $0.200 | 90% | 1 read |
Break-even assumes a cache write costs about 1.25x the base input rate, which is what OpenAI and Anthropic currently document for a short time-to-live. It answers one question: how many times a cached prefix must be re-read before caching is cheaper than paying full price each call. Above that count, every further read saves the discount shown.
Same input price, different output price
Both list $2 per million input tokens. Claude Sonnet 5 lists $10 per million output; GPT-5.6 Terra lists $12. On an output-heavy workload that 20 percent gap is the entire decision, and it is invisible if you compare input prices alone.
Reverse the shape and the answer reverses. A classification or extraction job that reads a lot and writes a little sees almost no difference between them.
These are the models most workloads should start on
Both sit in the balanced tier: roughly a fifth of frontier price, with a one-million-token context window. For most production tasks that is the right starting point, and moving up to a frontier model should follow a measured quality failure rather than a hunch.
The cheaper high-volume tier is a further order of magnitude down. GPT-5.6 Luna lists $0.20 input and $1.20 output, which is a tenth of Terra on input. If your task is routing, tagging, or summarising short text, test the cheap tier before assuming you need this one.
Long-context tier applies to one side only
GPT-5.6 Terra applies a long-context tier above 272,000 input tokens, doubling the input rate for the whole request. Claude Sonnet 5 bills its full window flat.
For a codebase-wide analysis or a long document, that single rule can matter more than the output-rate difference. The last two rows of the computed table show the effect directly.
Model the workload, not the marketing price.
Input price is identical, so compare output rate, cached-input rate, and long-context behaviour. At production volume these compound into a much larger difference than the sticker price suggests.
Review calculation methodology →Measure against the model you use
Other comparisons worth running
Compare approaches, not just models
Frequently asked questions
Is GPT-5.6 Terra or Claude Sonnet 5 cheaper?+
They are identical on input at $2 per million. Claude Sonnet 5 is cheaper on output at $10 per million against $12. For requests above 272,000 input tokens GPT-5.6 Terra applies a long-context tier, which widens the gap further.
Which should I use for a production agent?+
Test both on your real prompt. Claude Sonnet 5 has the lower output rate and the lower cached-input rate, which suits agent loops that resend a long prefix. GPT-5.6 Terra gives exact local token counts, which makes cost forecasting easier to verify.
Do I need a frontier model instead?+
Usually not at first. The balanced tier costs about a fifth of frontier pricing. Move up only after a measured quality failure on your own task, and compare the cheap tier first if your task is short-form.