THE TICKER · HEAD-TO-HEAD · CHECKED DAILY · LAST CHANGE 2026-08-14
CLAUDE OPUS 4.8 VS GPT-5.6 LUNA
Anthropic vs OpenAI, in USD per million tokens, cross-checked daily against the public pricing dataset. Same jobs, both meters running.
Claude Opus 4.8 costs 21.1x more than GPT-5.6 Luna for the same job — a 1,000-token prompt with a 3,000-token reply runs $0.08 vs $0.0038. At 1,000 runs a day, that gap is $2,286.00 a month for identical traffic.
SIDE BY SIDE
| Claude Opus 4.8 → | GPT-5.6 Luna → | |
|---|---|---|
| Vendor | Anthropic | OpenAI |
| Input / MTok | $5.00 | $0.200 |
| Output / MTok | $25.00 | $1.20 |
| Output multiple | 5.0x | 6.0x |
| 1K prompt + 3K reply | $0.08 | $0.0038 |
SAME JOB, BOTH BILLS
| Scenario | Claude Opus 4.8 | GPT-5.6 Luna | Gap |
|---|---|---|---|
| A typical prompt | $0.02 | $0.0011 | 21.1x |
| A 20-turn conversation | $0.12 | $0.0056 | 21.4x |
| Summarizing a 10-page document | $0.12 | $0.0054 | 22.1x |
| One agent run | $0.31 | $0.01 | 20.9x |
| A code review request | $0.10 | $0.0048 | 21.1x |
Output sizes use the same editorial ratios as the Burn Calculator. Estimates, not invoices — and price is not a benchmark: the cheaper model is only a bargain if it does the job.
SETTLE IT WITH YOUR OWN PROMPT
Paste a real prompt into the Burn Calculator and see what it costs on Claude Opus 4.8, GPT-5.6 Luna and the other 18 models on the board — plus how much of it is waste either way.