Moonshot AI narrows gap with Claude via new Kimi K3.1 Pro release

Moonshot AI has officially launched Kimi K3.1 Pro, a high-performing reasoning model that has climbed to the fourth spot on the Artificial Analysis Intelligence Index. The new system achieves a score of 55, positioning it closely behind the market leaders. This release marks a significant step for the Chinese developer in the global artificial intelligence race.

The primary claim is that K3.1 Pro represents a substantial leap in capability for an open-weights contender. It trails only Anthropic's Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full in overall rankings. If Moonshot releases the full model weights by early November, this update would become the highest-scoring open-weights model available. Such a move would narrow the performance divide between open and proprietary systems from twelve points to just three.

Performance gains are most evident in agentic tasks and coding. On the GDPval-AA benchmark, the model posts an Elo rating of 1,846, which is 312 points higher than its predecessor. In terminal-based coding tests, it reports a success rate of 65.7 percent, narrowly missing the top proprietary models. The company also notes a sharp decline in hallucination rates, dropping to 8.3 percent on their internal stress tests. While this is an improvement over the previous 23.4 percent, it still lags behind GPT-6 Astra, which records a 5.5 percent rate.

The model also demonstrates strong capabilities in 3D modeling and finance. It can interact directly with game engines like Unity and Unreal Engine without additional connectors. On the BenchCAD benchmark, it scores 92.7 percent with tools, approaching the performance of top competitors. For financial tasks, the system shows double-digit point increases in both modeling and research categories. It currently leads all tested models on the OfficeQA Pro v2 benchmark for general office work.

Despite the performance boost, the underlying architecture remains consistent. The model retains 2.8 trillion total parameters with 104 billion active ones and supports a one-million-token context window. However, the cost of using the API has dropped by 33 percent. Input costs are now $2 per million tokens, while output costs stand at $10. This pricing structure keeps the cost per intelligence index task nearly identical to the previous version, as the new model generates slightly more output tokens.

Moonshot has also addressed cybersecurity concerns in its technical documentation. The company found that vulnerability research abilities grew as post-training scaled. To prepare for the open-weights release, they removed specific cyber-related data from the training corpus. This adjustment reduced the model's ability to solve complex security tasks from 90.5 percent to 81.0 percent on internal benchmarks. The company aims to balance offensive capability with defensive utility for the broader open-source community.

A high-speed variant of the model is also available, running two to three times faster than the standard version. Both options are accessible through the company's various platforms, including Kimi.ai and the API. The release is expected to intensify competition among major AI providers as the November weight release date approaches.