GPT-5.6 Luna (High) vs Qwen3.8 27B (Medium)
Which is better for the money: GPT-5.6 Luna (High) by OpenAI or Qwen3.8 27B (Medium) by Alibaba? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-27.
Clear winner
Verdict
Pick GPT-5.6 Luna (High)
Qwen3.8 27B (Medium) vs GPT-5.6 Luna (High): 14.0% less intelligence, 1081.3% more cost to run the AA benchmark, and 241.9% slower end-to-end.
| Metric | GPT-5.6 Luna (High) | Qwen3.8 27B (Medium) | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 32.1 | 27.6 | −14.0% Qwen3.8 27B (Medium) is 14.0% less intelligent |
| Total AA cost (lower is better) | $108.15 | $1,277.59 | +1081.3% Qwen3.8 27B (Medium) is 12× more expensive |
| End-to-end time (lower is better) | 16.30s | 55.73s | +241.9% Qwen3.8 27B (Medium) is 241.9% slower |
| Price per 1M tokens (in / out) | $0.20 / $1.20 | $0.50 / $3.00 | |
| Input modalities | Text, Image | Text, Image, Video |
Today's Pareto frontier
Against the frontier
- GPT-5.6 Luna (High) is off the frontier: at the same or lower cost, Claude Haiku 5.5 (High) scores 37.80 for $94.1.
- Qwen3.8 27B (Medium) is off the frontier: at the same or lower cost, GPT-6.1 Sol (Max) scores 51.80 for $1082.
GPT-5.6 Luna (High)Qwen3.8 27B (Medium)Today's Pareto frontierOpen-weights frontierEvery other model
Matchup
Compare something else
Agents:
Accept: text/markdown or ?format=md returns this page as Markdown. JSON: /api/v1/compare/gpt-5-6-luna-high/qwen3-8-27b-medium · Markdown: /api/v1/compare/gpt-5-6-luna-high/qwen3-8-27b-medium?format=md · llms.txt · agent skill