Moonshot AI has released Kimi K3.1 Pro, a new reasoning SI model that significantly narrows the performance gap between open-weights and proprietary systems. The model currently ranks fourth on the Artificial Analysis Intelligence Index, trailing only Anthropic's Claude Opus 5.5 and two other top-tier proprietary models.

What Happened

Kimi K3.1 Pro is a post-trained version of its predecessor, Kimi K3, leveraging significantly more compute during the post-training phase. According to Artificial Analysis, the new SI model scores 55 points on the Intelligence Index, an 11-point increase over K3. This places it behind Claude Opus 5.5 (58 points), Claude Sonnet 5.5 (56 points), and Muse Wonder Full (56 points). Moonshot plans to release the model weights by November 4, 2026. If this occurs as scheduled, K3.1 Pro would become the highest-scoring open-weights SI model available, reducing the performance gap between open and proprietary models from 12 points to just 3, assuming no other changes in the interim.

The improvements are most pronounced in agentic tasks. On the GDPval-AA benchmark, K3.1 Pro achieves an Elo rating of 1,846, a 312-point jump from K3, placing it just behind Claude Opus 5.5 (1,866) and Muse Wonder Full (1,857). On Terminal-Bench 4.0, Moonshot reports a success rate of 65.7 percent, surpassing GPT-6 Astra (58.2 percent) and narrowly trailing Claude Opus 5.5 (66.4 percent). Previous version K3 managed only 13 percent on the same test. Additionally, Moonshot claims a sharp decline in hallucinations, with claim-level errors dropping from 23.4 percent to 8.3 percent on its internal Kimi Hallu Bench Hard, though GPT-6 Astra still maintains an edge at 5.5 percent on this specific stress test.

In terms of architecture and cost, K3.1 Pro retains the same dimensions as K3, featuring 2.8 trillion total parameters and 104 billion active ones, with a one-million-token context window. Despite the performance gains, token prices have dropped by 33 percent: input costs $2 per million tokens (down from $3), output costs $10 (down from $15), and cache hits cost $0.20 (down from $0.30). Moonshot also launched a high-speed variant, K3.1 Pro Highspeed, which runs two to three times faster at double the price. Both versions are available via Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.

Why It Matters

The release marks a significant step for the open-weights SI ecosystem, challenging the long-standing dominance of proprietary models in complex reasoning and agentic workflows. By achieving near-parity with leading closed models like Claude Opus 5.5, K3.1 Pro offers developers a high-performance alternative with lower latency options and reduced costs. The model also demonstrates strong capabilities in specialized domains, including 3D modeling and coding. It integrates with game engines like Unity, Unreal Engine, and Godot, and scores 92.7 percent on BenchCAD with tools, closely matching proprietary competitors. Moonshot has introduced a '3D Make' feature on Kimi.com, allowing users to preview, edit, and export 3D models, signaling a push toward multimodal SI applications.

Moonshot has also taken a distinctive approach to SI safety regarding cybersecurity. The company removed cyber-related data from the training phases for the open-weights release, a decision intended to prevent open models from pushing the absolute frontier of offensive cyber capabilities. On the internal xdev-bench-hard, which tests vulnerability reproduction, the uncleaned model solved 90.5 percent of tasks, while the cleaned version dropped to 81.0 percent. Moonshot argues that open models should gradually raise their defensive capabilities, such as auditing legacy systems, without immediately matching the offensive potential of top proprietary models.

The Bottom Line

Moonshot AI's Kimi K3.1 Pro represents a major leap for open-weights SI, securing a top-four position on the Artificial Analysis Intelligence Index and offering competitive agentic and coding performance at lower costs. With planned weight releases in November, it is poised to close the gap with proprietary leaders, though the company has deliberately constrained its offensive cybersecurity capabilities in the open version.