Claude Opus 4.8 vs GPT-5: API Pricing & Benchmarks
Claude Opus 4.8 vs GPT-5 — or, searched the other way round, GPT-5 vs Claude Opus 4.8: official API prices, what the same workload costs on each, and every published benchmark score, side by side. Every figure comes straight from the data below; this page does not pick a winner.
Price per 1M tokens
| Claude Opus 4.8 | GPT-5 | |
|---|---|---|
| Input | $5.00 | $1.25 |
| Output | $25.00 | $10.00 |
| Blended | $10.00 | $3.4375 |
| Developer | Anthropic | OpenAI |
| Released | 2026-05-28 | 2025-08-07 |
Prices: official list prices as recorded by models.dev (Claude Opus 4.8) and models.dev (GPT-5).
Cost of the same workload
| Workload | Tokens | Claude Opus 4.8 | GPT-5 | Difference |
|---|---|---|---|---|
| Short chat reply | 500 in / 300 out | $0.0100 | 0.362¢ | GPT-5 cheaper by 0.638¢ |
| RAG query | 4,000 in / 500 out | $0.0325 | $0.0100 | GPT-5 cheaper by $0.0225 |
| Long document summary | 50,000 in / 1,000 out | $0.2750 | $0.0725 | GPT-5 cheaper by $0.2025 |
| Code review of a 2k-line file | 30,000 in / 2,000 out | $0.2000 | $0.0575 | GPT-5 cheaper by $0.1425 |
| 1,000 chat replies/day for 30 days | 30,000 × (500 in / 300 out) | $300.00 | $108.75 | GPT-5 cheaper by $191.25 |
Cost = (input tokens ÷ 1,000,000 × input price) + (output tokens ÷ 1,000,000 × output price), using each model's own prices above.
Benchmarks
| Benchmark | Claude Opus 4.8 | GPT-5 |
|---|---|---|
| Epoch Capabilities Index (ECI) | 159.9 epoch · as of 2026-05-28 |
150.0 epoch · as of 2025-08-07 |
| Epoch Mock AIME accuracy | 98.3% epoch · as of 2026-05-28 |
No public data yet |
| Epoch FrontierMath (T1–3) accuracy | 80.0% epoch · as of 2026-05-28 |
No public data yet |
| Arena text rating | 1474 arena · as of 2026-10-28 |
1435 arena · as of 2026-10-28 |
| Arena coding rating (human preference votes, not task accuracy) | 1529 arena · as of 2026-10-28 |
1471 arena · as of 2026-10-28 |
| Arena math rating | 1477 arena · as of 2026-10-28 |
1435 arena · as of 2026-10-28 |
| Arena WebDev rating | 1536 arena · as of 2026-10-28 |
1417 arena · as of 2026-10-28 |
Full pages
Claude Opus 4.8 pricing GPT-5 pricing Claude Opus 4.8 token counter GPT-5 token counter
FAQ
Which is cheaper, Claude Opus 4.8 or GPT-5?
GPT-5 is cheaper on blended price: $3.4375 vs $10.00 per 1M tokens, which is $6.5625 less per 1M tokens (65.6% below Claude Opus 4.8).
How do the input and output prices of Claude Opus 4.8 and GPT-5 compare?
Claude Opus 4.8 costs $5.00 per 1M input tokens and $25.00 per 1M output tokens. GPT-5 costs $1.25 per 1M input tokens and $10.00 per 1M output tokens.
What are the ECI scores of Claude Opus 4.8 and GPT-5?
Claude Opus 4.8 scores 159.9 on the Epoch Capabilities Index (ECI); GPT-5 scores 150.0 on the Epoch Capabilities Index (ECI).
Where does the Claude Opus 4.8 vs GPT-5 comparison data come from?
Prices come from models.dev and LiteLLM (open MIT-licensed catalogs); benchmark scores come from Epoch AI and LMArena (open CC BY 4.0 datasets). This page states the numbers side by side and does not pick an overall winner — see /data-sources/.