Moonshot's Kimi K3.1 Pro lands fourth overall at a third of the price. The frontier gap is now three points

The open-weight frontier moved to three points on Friday, and the price did most of the talking.

Moonshot released Kimi K3.1 Pro, a 2.8 trillion parameter MoE that scored 55 on the Artificial Analysis Intelligence Index, fourth of 215 models and three points behind Claude Opus 5.5, per the index coverage.

The price: $2 per million input tokens and $10 per million output, roughly a third below its predecessor, the release tracking said.

Weights promise lands November 4, so the closed preview carries an open deadline.

The result marks an 11-point jump over Kimi K3, the highest score yet for a model whose weights are promised as open.

What three points prices

Follow the money at the gap number, since it decides who switches.

Claude Opus 5.5 carries a frontier rate behind Anthropic's own tier structure, while K3.1 Pro matches within three points at $2 and $10. Three points of index for a fraction of the bill changes procurement rubrics more than any launch note this month.

Under the hood: the open-weight promise carries the extra risk story.

An open release due November 4 lets every team run, verify, and self-host the same result the index measured, so the score stops being a vendor claim the day the weights land.

Ecosystem check, in the month's block: Mistral pushed its trillion-parameter open model earlier in the month, Chinese releases carried the week cadence, and the open tier now sits inside four points of the closed frontier.

Precedent says the pattern compounds: China's Kimi was the model that owed its own IPO filing to this demand curve, so the market carries both sides of the same bet.

One caveat worth carrying: 55 sits with Chinese operators' published scores, and independent reruns matter once the weights ship.

I get the sense frontier products now carry a definitional problem. What stays worth paying for inside four points runs thin unless the product owns the agents.

Source