Is Mistral Large 4 anywhere near the frontier?

Short answer: no. Longer answer: it depends on what you are measuring, and the scoreboard is brutal.

On general smarts, not close. The Artificial Analysis Intelligence Index gives Large 4 a 38. The Chinese open models GLM-5.3 and Kimi K3 score 45 and 44, and the top closed models, Claude Opus 5.5 and GPT-6 Astra, are further out of reach still. A second tracker, Vals AI, ranks it around 32nd or 33rd out of roughly 44 to 45 models, at about 48 percent on its index.

So is it just a bad model that got a lot of press?

No, and that is the interesting part. A 38 sits well above the median of 26 among comparable models on that same index. On a benchmark for legal agents it ranks sixth out of 76. This is a good model in a league where "good" is no longer enough.

That is the real story of the frontier right now. You can build something strong, spend a mountain of money, and still land in the middle of the table, because the top keeps sprinting away.