AdvertiseDonate
Latest20 Labour MPs Pay 9,600 Pounds for Leadership Training
ChicagoMilwaukeeMadisonFort WayneIndianapolisToledoDetroitCincinnati
News Locall · 330 North Wabash Avenue Suite 23rd Floor, Chicago, IL 60611
Business · AI

Moonshot AI cuts Kimi K3.1 Pro prices by 33% to close gap with Claude Opus 5.5

The new Chinese model jumps 11 points on intelligence rankings while reducing API costs, aiming to dominate the open-weights market by November.

By Adrian Romano ·

Illustration for: Moonshot AI cuts Kimi K3.1 Pro prices by 33% to close gap with Claude Opus 5.5
Illustration generated for this article, not a news photograph.

55 points. That is the score Moonshot AI’s latest model, Kimi K3.1 Pro, achieved on the Artificial Analysis Intelligence Index, placing it just three points behind the industry-leading proprietary systems. This performance metric signals a significant shift in the competitive landscape for large language models, as the Chinese firm narrows the efficiency and capability gap with Western rivals.

The release follows a clear trajectory of rapid iteration. Moonshot previously launched Kimi K3, which served as the baseline for the current version. By applying significantly more post-training compute, the company has effectively upgraded the architecture without altering its fundamental size. The result is a model that not only matches top-tier competitors in general reasoning but also shows substantial gains in specialized agentic tasks and coding environments.

The financial implications of this update are immediate and tangible for developers. While the parameter count remains unchanged at 2.8 trillion total, with 104 billion active parameters, the pricing structure has been aggressively adjusted. Input costs have dropped from $3 to $2 per million tokens, while output costs fell from $15 to $10. Cache hits are now priced at $0.20, down from $0.30. These cuts represent a 33 percent reduction in overall API expenses, making the model highly attractive for high-volume enterprise deployments.

Performance data underscores the value proposition. On the GDPval-AA benchmark, K3.1 Pro reached 1,846 Elo, a 312-point increase over its predecessor. It now trails only Claude Opus 5.5 and Muse Wonder Full in that specific category. In terminal-based coding tasks, the new model scored 65.7 percent on Terminal-Bench 4.0, surpassing GPT-6 Astra and coming within less than a point of Claude Opus 5.5. Meanwhile, hallucination rates on internal stress tests dropped sharply from 23.4 percent to 8.3 percent, indicating a more reliable output stream for business applications.

Beyond core reasoning, the model demonstrates strong capabilities in 3D modeling and game development. It can interact directly with engines like Unity and Unreal Engine. On the BenchCAD benchmark, it achieved 92.7 percent accuracy with tools enabled, closely matching top competitors. This technical versatility is supported by a new feature on Kimi.com called "3D Make," which allows users to preview and export models, expanding the model's utility beyond text generation into spatial design.

The strategic move extends to the open-weights market. Moonshot plans to release the model weights by November 4, 2026. If executed as planned, K3.1 Pro would become the highest-scoring open-weights model available. This release would shrink the performance gap between open and proprietary models from 12 points to just three, fundamentally altering the cost-benefit analysis for organizations weighing licensing fees against self-hosting capabilities.

To prepare for the open-weights release, Moonshot implemented deliberate limitations on offensive cybersecurity capabilities. By removing specific cyber-related data from the training corpus, the company reduced the model's score on an internal vulnerability reproduction test from 90.5 percent to 81.0 percent. This decision reflects a broader industry debate over the balance between offensive power and safety in open models, acknowledging that while these tools serve defensive auditing roles, they should not necessarily define the absolute frontier of offensive capabilities.

For the market, the arrival of K3.1 Pro represents a consolidation of value. With the same parameter efficiency as its predecessor but lower costs and higher scores, it sits at the top of the Pareto frontier for its size. The availability of both a standard version and a "Highspeed" variant, which runs two to three times faster, provides businesses with flexible options to optimize for either depth of reasoning or latency. As the November deadline approaches, the focus will shift to how quickly the open-weights release can capture market share from established proprietary leaders.

More Business · Watch Your Wallet · Report a business · Got a Story?

Get News Locall in your inboxLocal news, all in one place. Free, a few times a week, unsubscribe any time.

Related stories

Get News Locall in your inboxLocal news, all in one place. Free, a few times a week, unsubscribe any time.