Chart comparing AI models: Kimi K3.1 Pro leading OfficeQA Pro v2, both approaching Claude Opus 5.5.

Editorial illustration for Kimi K3.1 Pro tops OfficeQA Pro v2, closing in on Anthropic's Claude Opus 5.5.

Kimi K3.1 Pro tops OfficeQA Pro v2, closing in on...

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

October 23, 2026 • 3 min read

Moonshot AI put a number on how close Chinese open-weight models have gotten to the best American systems money can buy. The company's new Kimi K3.1 Pro scored 55 points on the Artificial Analysis Intelligence Index, an 11-point jump from its predecessor K3, and landed in fourth place overall. Only three models sit above it: Claude Opus 5.5 at 58, Claude Sonnet 5.5 at 56, and Muse Wonder Full at 56.

The jump comes from extra post-training compute applied to the same 2.8 trillion parameter base, with Moonshot reporting gains in agentic tasks, coding, and 3D modeling. API pricing dropped 33 percent at the same time. Moonshot also shipped a faster variant, K3.1 Pro Highspeed, running two to three times quicker, and made both versions available immediately through Kimi.ai, Kimi Work, and related platforms.

The bigger story is timing. Moonshot has committed to releasing the model's weights by November 4, 2026. If that holds and nothing else shifts in the meantime, K3.1 Pro would become the strongest open-weights model on the market, cutting the distance between open and closed systems down to a handful of points.

If Moonshot releases the model weights as planned by November 4, 2026, K3.1 Pro would become the highest-scoring open-weights model available. That would shrink the gap between open and proprietary models from 12 to just 3 points, assuming nothing else changes by then.

Chinese AI model Kimi K3.1 Pro is closing in on Anthropic's Claude Opus 5.5 - THE DECODER

Why this matters A fourth-place finish on one index, built on benchmarks Moonshot itself ran and reported, isn't the same as beating Claude Opus 5.5 in production. We'd want to see independent verification of those OfficeQA Pro v2 numbers before treating this as a genuine lead change, especially since the 7.2 and 11.6 point gains on finance benchmarks come from Moonshot's own internal tests with reasoning effort cranked to "max," a setting that tends to flatter demo results more than everyday usage. Still, the real story for builders might be the price cut.

A 33 percent drop in API costs on a 2.8 trillion parameter model, paired with a promise to open the weights by November 2026, is the kind of move that pressures every lab competing on agentic and coding tasks, not just Anthropic. For founders scoping products around office automation or financial research agents, Kimi K3.1 Pro is now a line item worth pricing against GPT and Claude, not a curiosity. For researchers, the open-weights timeline matters more than the leaderboard rank: if Moonshot follows through, it changes who gets to study a frontier-scale model up close.