Claude Haiku 5.5 (Xhigh) vs GPT-6 Astra (Low)
Which is better for the money: Claude Haiku 5.5 (Xhigh) by Anthropic or GPT-6 Astra (Low) by OpenAI? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-27.
Trade-off
Verdict
Depends on your budget
GPT-6 Astra (Low) scores 4.6 points higher (45.8 vs 41.2) but costs 9.8× as much to run the benchmark ($1537 vs $158; premium vs mid-range cost classes). They are not alternatives at one budget: pick GPT-6 Astra (Low) when you need intelligence above 41.2, and Claude Haiku 5.5 (Xhigh) when 41.2 is enough.
| Metric | Claude Haiku 5.5 (Xhigh) | GPT-6 Astra (Low) | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 41.2 | 45.8 | +11.2% GPT-6 Astra (Low) is 11.2% more intelligent |
| Total AA cost (lower is better) | $157.58 | $1,536.78 | +875.2% GPT-6 Astra (Low) is 9.8× more expensive |
| End-to-end time (lower is better) | 75.35s | 12.08s | −84.0% GPT-6 Astra (Low) is 84.0% faster |
| Price per 1M tokens (in / out) | $0.10 / $0.50 | $10.0 / $50.0 | |
| Input modalities | Text, Image | Text, Image |
GPT-6 Astra (Low) vs Claude Haiku 5.5 (Xhigh): 11.2% more intelligence, 875.2% more cost to run the AA benchmark, and 84.0% faster end-to-end.
Today's Pareto frontier
Against the frontier
- Claude Haiku 5.5 (Xhigh) is on today's frontier: nothing is both smarter and cheaper.
- GPT-6 Astra (Low) is off the frontier: at the same or lower cost, GPT-6.1 Sol (Max) scores 51.80 for $1082.
Claude Haiku 5.5 (Xhigh)GPT-6 Astra (Low)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/claude-haiku-5-5-xhigh/gpt-6-astra-low · Markdown: /api/v1/compare/claude-haiku-5-5-xhigh/gpt-6-astra-low?format=md · llms.txt · agent skill