Moonshot AI slashes Kimi K3.1 Pro costs as it narrows gap with Anthropic

Moonshot AI has launched Kimi K3.1 Pro, a high-performance reasoning model that significantly reduces processing costs while improving benchmark scores. The release positions the Chinese firm to challenge dominant proprietary systems from Anthropic and OpenAI. By lowering API prices by 33 percent, the company aims to make advanced AI capabilities more accessible to developers without sacrificing performance metrics.

The new model scores 55 points on the Artificial Analysis Intelligence Index. This result marks an 11 point increase over its predecessor, Kimi K3. It currently ranks fourth, trailing only Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full. If Moonshot releases the model weights by November 4, 2026, K3.1 Pro would become the highest scoring open weights model available. Such a move would shrink the performance gap between open and proprietary models from 12 points to just 3 points.

Performance gains are most visible in agentic tasks and coding. On the GDPval-AA benchmark, the model achieves 1,846 Elo, a 312 point leap from the previous version. It also records 65.7 percent on Terminal-Bench 4.0, surpassing GPT-6 Astra and closely following Claude Opus 5.5. Hallucination rates have dropped sharply as well. Internal tests show claim level hallucinations falling from 23.4 percent to 8.3 percent. However, GPT-6 Astra maintains a slight edge with a 5.5 percent rate on the same stress test.

The model also excels in 3D modeling and office productivity. It integrates directly with game engines like Unity and Unreal Engine. On BenchCAD, it scores 92.7 percent with tools enabled. Moonshot states the system is strongest at parametric modeling of mechanical parts and architectural structures. A new feature called 3D Make now allows users to preview and export 3D models directly on Kimi.com. For general office tasks, the model tops all tested competitors on OfficeQA Pro v2.

Pricing reflects a strategic shift toward volume and efficiency. Input tokens now cost $2 per million, down from $3. Output tokens drop to $10 from $15, and cache hits fall to $0.20. The total cost per Intelligence Index task remains stable at $1.98, nearly identical to the previous model. This stability occurs because the new version generates 14 percent more output tokens, offsetting the lower per token rate. The architecture retains 2.8 trillion total parameters with 104 billion active, supporting a one million token context window.

Moonshot also addressed cybersecurity concerns in its technical documentation. The company noted that vulnerability research capabilities grew as post training scaled. To manage this, cyber related data was stripped from the training data for the open weights release. The cleaned model solves 81.0 percent of tasks on an internal benchmark, compared to 90.5 percent for the uncleaned version. This approach balances offensive capability with defensive utility for auditing legacy systems.

The launch includes a high speed variant that runs two to three times faster. Both versions are available immediately through Kimi.ai, Kimi Work, and the Kimi API. The move signals an aggressive push by Chinese AI firms to close the technological divide with US based rivals through price competition and rapid iteration.