Moonshot AI's Kimi K3.1 Pro Narrows Gap With Leading US Rivals
A new release from a Chinese firm lands in fourth place on a major intelligence benchmark, challenging the dominance of top Western models while lowering costs for developers.
By Olivia Miller ·
The race for artificial intelligence supremacy has traditionally been a two-horse competition between major American laboratories and a handful of strong European players, but a new entrant from the East is reshaping that narrative. This week, Moonshot AI released Kimi K3.1 Pro, a significant upgrade to its previous model that immediately placed it in the upper tier of global performance metrics.
The new model achieved a score of 55 on the Artificial Analysis Intelligence Index, marking an 11-point improvement over its predecessor, Kimi K3. This performance secures K3.1 Pro the fourth position on the leaderboard. It trails only Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full, which currently hold the top three spots. If Moonshot follows through on its plan to release the model weights by November 4, 2026, K3.1 Pro will become the highest-scoring open-weights model available. Such a move would drastically reduce the performance gap between open and proprietary systems from 12 points to just three, assuming no other major releases occur in the interim.
Performance gains are most visible in complex, agentic workflows. On the GDPval-AA benchmark, the new model reached an Elo score of 1,846, surpassing its predecessor by 312 points and placing it just behind the leading proprietary models. In coding tasks, specifically on Terminal-Bench 4.0, K3.1 Pro recorded a success rate of 65.7 percent. This figure exceeds the performance of GPT-6 Astra and comes within a fraction of a percent of Claude Opus 5.5. Notably, the earlier Kimi K3 model managed only 13 percent on the same test, highlighting the rapid evolution in the company's engineering approach.
Beyond raw capability, Moonshot has focused on reliability and cost efficiency. The company reports that hallucination rates have dropped significantly, with claim-level errors falling from 23.4 percent to 8.3 percent on an internal stress test. While this does not represent general error rates, it indicates a substantial reduction in factual inaccuracies. Financially, the API pricing has been reduced by 33 percent across the board. Input tokens now cost $2 per million, down from $3, while output tokens are priced at $10, a drop from $15. Despite these price cuts, the model maintains the same architectural dimensions as its predecessor, featuring 2.8 trillion total parameters and a one-million-token context window.
The release also targets specialized applications in 3D modeling and cybersecurity. K3.1 Pro can interact directly with game engines like Unity and Unreal Engine, scoring 92.7 percent on the BenchCAD benchmark when using tools. Moonshot has also introduced a feature called 3D Make on its platform, allowing users to preview and export models. On the security front, the company took deliberate steps to limit offensive capabilities in the open-weights version by removing specific cyber-related data from training. This adjustment lowered the model's success rate on a difficult vulnerability reproduction task from 90.5 percent to 81.0 percent, a trade-off the company argues is necessary to balance offensive power with defensive utility for auditing legacy systems.
A high-speed variant of the model, named K3.1 Pro Highspeed, is also available now. It runs two to three times faster than the standard version, offering developers a choice between maximum accuracy and rapid execution. Both versions are accessible through Kimi.ai, Kimi Work, Kimi Code, and the standard API. For Texas and South Central businesses relying on cloud-based AI services, the combination of lower prices and improved agentic performance offers a compelling alternative to the more expensive proprietary options currently dominating the market.
