Mistral Large 3 vs Mistral Large 4 Preview
Which is better for the money: Mistral Large 3 by Mistral or Mistral Large 4 Preview by Mistral? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-24.
Trade-off
Verdict
Depends on your budget
Mistral Large 4 Preview scores 29.1 points higher (38.4 vs 9.3) but costs 35× as much to run the benchmark ($1608 vs $46.4; premium vs budget cost classes). They are not alternatives at one budget: pick Mistral Large 4 Preview when you need intelligence above 9.3, and Mistral Large 3 when 9.3 is enough.
| Metric | Mistral Large 3 | Mistral Large 4 Preview | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 9.3 | 38.4 | +312.9% Mistral Large 4 Previ… is 4.1× more intelligent |
| Total AA cost (lower is better) | $46.41 | $1,607.86 | +3364.5% Mistral Large 4 Previ… is 35× more expensive |
| End-to-end time (lower is better) | 7.18s | 27.12s | +277.7% Mistral Large 4 Previ… is 277.7% slower |
| Price per 1M tokens (in / out) | $0.50 / $1.50 | $1.36 / $4.18 | |
| Input modalities | Text, Image | Text, Image |
Mistral Large 4 Preview vs Mistral Large 3: 312.9% more intelligence, 3364.5% more cost to run the AA benchmark, and 277.7% slower end-to-end.
Today's Pareto frontier
Against the frontier
- Mistral Large 3 is off the frontier: at the same or lower cost, GPT-6 Luna (Medium) scores 29.90 for $31.2.
- Mistral Large 4 Preview is off the frontier: at the same or lower cost, GPT-6.1 Sol (Max) scores 51.80 for $1082.
Mistral Large 3Mistral Large 4 PreviewToday'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/mistral-large-3/mistral-large-4 · Markdown: /api/v1/compare/mistral-large-3/mistral-large-4?format=md · llms.txt · agent skill