Comparison · verified August 21, 2026
Claude vs GPT-5.6 for Customer Support
Support is the one application where the cheapest capable model usually wins, because volume dominates everything else. It is also the application where a confidently wrong answer costs the most, because it reaches a customer with your name on it.
Those two facts pull in opposite directions, and that tension — not benchmark scores — is what should decide this. Below: verified pricing at real ticket volume, the behavioural differences that matter for escalation, and a shadow-mode test to run before either model touches a live conversation.
The verified differences
What each one costs
Last verified August 21, 2026
USD per million tokens, standard rates, no discounts applied. Read the three caveats underneath before you put these numbers in a spreadsheet.
| Model | Tier | Input | Output | Notes |
|---|---|---|---|---|
| Claude Fable 5 | frontier | $10 | $50 | The frontier model. Twice the price of Opus 5 for the hardest work. |
| Claude Opus 5 | frontier | $5 | $25 | The default flagship. Thinking on by default. Fast mode available at $10/$50. |
| Claude Sonnet 5 | balanced | $2 | $10 | What most production agent fleets run on. The September 2026 increase to $3/$15 was cancelled. |
| Claude Haiku 4.5 | fast | $1 | $5 | High-volume and latency-sensitive work. |
| GPT-5.6 Sol | frontier | $5 | $30 | The flagship tier. Long-context requests meter at $10/$45. |
| GPT-5.6 Terra | balanced | $2 | $12 | The everyday tier, cut to this rate on July 30, 2026. Long-context meters at $4/$18. |
| GPT-5.6 Luna | fast | $0.20 | $1.2 | The cheapest frontier-family option on the market. Long-context meters at $0.40/$1.80. |
Anthropic changed tokenizers at Claude 4.7
Claude 4.7 and later — including Opus 5, Sonnet 5, and Fable 5 — use a newer tokenizer that produces roughly 30% more tokens for the same text than Sonnet 4.6 and earlier. A $2/MTok model on the new tokenizer is not directly comparable to a $2/MTok model on an older one, or to another vendor. Compare cost per task, not cost per token.
OpenAI meters long context separately
GPT-5.6 publishes a second, higher rate for long-context requests: Sol goes from $5/$30 to $10/$45, Terra from $2/$12 to $4/$18, Luna from $0.20/$1.20 to $0.40/$1.80. Anthropic includes the full 1M-token window at standard pricing on Claude 4.6 and later — a 900k-token request costs the same per token as a 9k one. If your workload is context-heavy, that difference is larger than the headline gap.
Two of Google’s current rates are introductory
Gemini 3.7 Flash and 3.6 Flash are priced at $0.75/$3.75 only through December 31, 2026. Both double to $1.50/$7.50 in 2027. If you are building a twelve-month cost model on Gemini Flash, model the 2027 number, not the one on the page today.
Caching and batching move the number more than model choice
A Claude cache hit costs 10% of the standard input price, and the Batch API takes 50% off both directions; the two stack. A workload that reuses a large system prompt can land below a nominally cheaper model that you are calling uncached. Model selection is usually the third-largest lever, after caching and after not sending the context at all.
What the evidence says, and who is saying it
Almost every performance number in public circulation traces back to a vendor running its own harness. That does not make the numbers useless, but the attribution belongs next to the claim.
Anthropic, published cost guidance
- A worked example of 10,000 support conversations on Haiku 4.5 at roughly 3,700 tokens each totals about $37 in tokens.
- Cache reads are billed at 10% of the standard input rate, and the Batch API takes 50% off both directions.
These are Anthropic’s own figures for an illustrative workload. Your token-per-conversation number is the variable that matters, and it is usually higher than you expect once policy documents and conversation history are in context.
What neither vendor publishes
- Deflection rate on your ticket mix.
- How often the model escalates when it should have answered, and answered when it should have escalated.
- CSAT on AI-handled tickets versus human-handled ones.
These are the only three numbers that decide whether support automation works, and no vendor can produce them for you. They come out of a shadow-mode run, which is why the section below exists.
Run it in shadow mode before a customer sees it
Support is the wrong place to learn from production. The good news is that support is unusually easy to test safely, because you have a backlog of resolved tickets with known correct answers and a human already graded them.
- Take 200 resolved tickets from the last quarter. Weight them the way your real queue is weighted, not toward the interesting ones. The boring repetitive tickets are where the ROI lives.
- Run both models against them with no customer in the loop. Same prompt, same retrieved context, same tools. Log the answer and the escalate/answer decision.
- Grade three things separately. Was the answer correct? Was it grounded in a real source passage? Did it make the right escalate-or-answer call? A model can score well on the first and badly on the third, and the third is what gets you in trouble.
- Count the confidently wrong answers specifically. Not the refusals — the answers that were fluent, plausible, and wrong. That number, not the average, is your risk. One wrong refund policy quote outweighs fifty good deflections.
- Price it on cost per resolved ticket. Include the escalated ones, which cost you tokens and a human. A model that is cheaper per token and escalates twice as often is not cheaper.
How to evaluate your agents has the test-set design, and hallucination prevention covers the grounding setup that keeps the confidently-wrong number down.
The honest bottom line
Pick Claude if the cost of a wrong answer is high — regulated industries, financial or medical information, anything where a customer could act on bad guidance. The citation behaviour and the readier escalation are worth real money in those settings, and the compliance surface is more developed.
Pick GPT-5.6 if you are running very high volume on low-stakes tickets, where Luna at $0.20 / $1.20 changes the unit economics outright, or if your helpdesk already ships a first-class OpenAI integration and building the equivalent is a quarter of engineering time you do not have.
Route between them which is what most mature support deployments end up doing. Cheap model for tier-one deflection, stronger model for anything touching policy, money, or a frustrated customer. The routing rule is the product; the model is a component. See Claude for customer support for the full build.