GPT-6 Astra (Medium) vs Mercury 2.5
Which is better for the money: GPT-6 Astra (Medium) by OpenAI or Mercury 2.5 by Inception? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-25.
Trade-off
Verdict
Depends on your budget
GPT-6 Astra (Medium) scores 37.3 points higher (49.6 vs 12.3) but costs 11× as much to run the benchmark ($2434 vs $218; flagship vs mid-range cost classes). They are not alternatives at one budget: pick GPT-6 Astra (Medium) when you need intelligence above 12.3, and Mercury 2.5 when 12.3 is enough.
| Metric | GPT-6 Astra (Medium) | Mercury 2.5 | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 49.6 | 12.3 | −75.2% Mercury 2.5 is 75.2% less intelligent |
| Total AA cost (lower is better) | $2,434.12 | $218.11 | −91.0% Mercury 2.5 is 91.0% cheaper |
| End-to-end time (lower is better) | 13.89s | 4.18s | −69.9% Mercury 2.5 is 69.9% faster |
| Price per 1M tokens (in / out) | $10.0 / $50.0 | $0.25 / $0.75 | |
| Input modalities | Text, Image | Text |
Mercury 2.5 vs GPT-6 Astra (Medium): 75.2% less intelligence, 91.0% less cost to run the AA benchmark, and 69.9% faster 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.
- Mercury 2.5 is off the frontier: at the same or lower cost, MiMo-V2.6-Pro scores 46.30 for $207.
GPT-6 Astra (Medium)Mercury 2.5Today'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/mercury-2-5 · Markdown: /api/v1/compare/gpt-6-astra-medium/mercury-2-5?format=md · llms.txt · agent skill