Moonshot AI Slashes Kimi K3.1 Pro Pricing While Closing Gap on Anthropic

Moonshot AI has entered the high-performance reasoning market with the launch of Kimi K3.1 Pro, a model that secures the fourth spot on the Artificial Analysis Intelligence Index with a score of 55. This places the new release just three points behind the leading proprietary offerings, specifically Anthropic’s Claude Opus 5.5. For investors and procurement teams, the release is notable for its aggressive pricing strategy and its position on the efficiency frontier, offering high capability without a corresponding increase in infrastructure costs.

The financial impact of the update is immediate and significant. While the model maintains the same architectural dimensions as its predecessor, Kimi K3, with 2.8 trillion total parameters, Moonshot has reduced API pricing by 33 percent. Input costs have dropped to $2 per million tokens, output costs to $10, and cache hits to $0.20. This price reduction, combined with an 11-point jump in benchmark performance over the previous version, suggests a strategy aimed at capturing market share from both lower-tier models and expensive proprietary alternatives.

Performance gains are most pronounced in agentic workflows, which are critical for enterprise automation. On the GDPval-AA benchmark, Kimi K3.1 Pro achieved an Elo rating of 1,846, a 312-point increase over K3. This performance trails only Claude Opus 5.5 and Muse Wonder Full. In coding tasks, the model recorded a score of 65.7 percent on Terminal-Bench 4.0, surpassing GPT-6 Astra and approaching the top-tier proprietary leader. These metrics indicate that the model is viable for complex, multi-step software development tasks that previously required more expensive solutions.

The release also includes a variant named K3.1 Pro Highspeed, which the company claims runs two to three times faster than the standard version. Both models are currently available through Kimi.ai, Kimi Work, Kimi Code, and the standard API. The inclusion of a high-speed option addresses the latency concerns often associated with large parameter models, making the system more suitable for real-time applications where response time is a key component of the user experience.

A key differentiator for this release is its capability in 3D modeling and game engine integration. Moonshot reports that the model can directly interact with Unity, Unreal Engine, and Godot, handling tasks ranging from bug fixes to full project builds. It scored 92.7 percent on the BenchCAD benchmark when using tools, a result that is competitive with top proprietary models. The company has also launched a "3D Make" feature that allows users to preview and export models, expanding the product’s utility beyond traditional text and image generation.

From a data integrity perspective, the company has made specific adjustments to the model’s training data to balance offensive and defensive cybersecurity capabilities. By removing cyber-related data from the open-weights version, Moonshot aims to prevent the model from becoming too specialized in vulnerability exploitation. While this reduces its score on internal security benchmarks from 90.5 percent to 81.0 percent, the company argues that this level of capability is sufficient for auditing legacy systems while maintaining a safety margin for open-source deployments.

Looking ahead, the competitive dynamics of the AI market may shift depending on the upcoming release of Kimi K3.1 Pro’s model weights. If released as planned by November 4, 2026, it would become the highest-scoring open-weights model available. This event would narrow the performance gap between open and proprietary models from 12 points to just three, potentially altering the cost-benefit analysis for companies deciding between licensing proprietary APIs and deploying open-weights infrastructure.

The launch positions Moonshot as a serious contender in the mid-to-high tier of the reasoning model market. By combining near-frontier performance with reduced costs and expanded 3D capabilities, the company is targeting a segment of the market that previously relied exclusively on US-based competitors. As the weights become available, the market will likely see a reevaluation of the total cost of ownership for AI-driven development and automation projects.