GPT-6 Astra (Medium) vs GPT-6 Luna (High)
Which is better for the money: GPT-6 Astra (Medium) by OpenAI or GPT-6 Luna (High) by OpenAI? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-26.
Trade-off
Verdict
Depends on your budget
GPT-6 Astra (Medium) scores 16.7 points higher (49.6 vs 32.9) but costs 51× as much to run the benchmark ($2434 vs $47.9; flagship vs budget cost classes). They are not alternatives at one budget: pick GPT-6 Astra (Medium) when you need intelligence above 32.9, and GPT-6 Luna (High) when 32.9 is enough.
| Metric | GPT-6 Astra (Medium) | GPT-6 Luna (High) | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 49.6 | 32.9 | −33.7% GPT-6 Luna (High) is 33.7% less intelligent |
| Total AA cost (lower is better) | $2,434.12 | $47.85 | −98.0% GPT-6 Luna (High) is 98.0% cheaper |
| End-to-end time (lower is better) | 14.16s | 18.94s | +33.8% GPT-6 Luna (High) is 33.8% slower |
| Price per 1M tokens (in / out) | $10.0 / $50.0 | $0.10 / $0.50 | |
| Input modalities | Text, Image | Text, Image |
GPT-6 Luna (High) vs GPT-6 Astra (Medium): 33.7% less intelligence, 98.0% less cost to run the AA benchmark, and 33.8% slower end-to-end.
Today's Pareto frontier
Against the frontier
- GPT-6 Astra (Medium) is off the frontier: at the same or lower cost, Claude Opus 5.5 (High, Default Fallback) scores 53.60 for $2172.
- GPT-6 Luna (High) is on today's frontier: nothing is both smarter and cheaper.
GPT-6 Astra (Medium)GPT-6 Luna (High)Today's Pareto frontierEvery other model
Matchup
Compare something else
Agents:
Accept: text/markdown or ?format=md returns this page as Markdown. JSON: /api/v1/compare/gpt-6-astra-medium/gpt-6-luna-high · Markdown: /api/v1/compare/gpt-6-astra-medium/gpt-6-luna-high?format=md · llms.txt · agent skill