Moonshot AI's Kimi K3.1 Pro Ranks Fourth on Intelligence Index
A new Chinese model jumps 11 points to sit just behind top proprietary systems while cutting API costs.
By Isaac Medina ·
Moonshot AI has released Kimi K3.1 Pro, and it now sits in fourth place on the Artificial Analysis Intelligence Index. The new model scores 55 points. It is only three points behind the leaders.
The performance jump is significant. The model gained 11 points over its predecessor, Kimi K3. Only three models rank higher right now. Claude Opus 5.5 leads with 58 points. Claude Sonnet 5.5 and Muse Wonder Full both score 56. If Moonshot releases the model weights by November 4, 2026, as planned, it would become the top open-weights model. That move would shrink the gap between open and proprietary systems from 12 points to just three.
The company is also shipping a variant called K3.1 Pro Highspeed. This version runs two to three times faster. Both options are available now on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. The parameter count remains at 2.8 trillion total with 104 billion active. The context window covers one million tokens. Input prices drop to $2 per million tokens. Output costs fall to $10. Cache hits cost $0.20. These figures represent a 33 percent price cut.
The biggest improvements appear in agentic tasks. On the GDPval-AA benchmark, the model reaches 1,846 Elo. That is 312 points above K3. It trails only Claude Opus 5.5 and Muse Wonder Full. On Terminal-Bench 4.0, Moonshot reports a score of 65.7 percent. This beats GPT-6 Astra. It is also very close to Claude Opus 5.5. Hallucination rates dropped sharply as well. Internal tests show claim-level errors falling from 23.4 percent to 8.3 percent.
The model also shows strong results in 3D modeling and coding. It works with game engines like Unity and Unreal. On the BenchCAD test, it scores 92.7 percent with tools. This is close to the top models. Moonshot says it is strongest at parametric modeling of mechanical parts. A new feature called 3D Make lets users preview and export models on the website.
Moonshot is taking cybersecurity seriously for this release. The company stripped cyber-related data from its training sets. This is a deliberate limit on offensive capabilities. In internal tests, the uncleaned version solved 90.5 percent of vulnerability tasks. After cleaning, the score is 81.0 percent. The company argues that open models should not push the absolute frontier of offensive power. However, it wants to gradually raise what they can do for defensive purposes.
The stakes are clear for the industry. The gap between free open models and paid proprietary ones is closing fast. Costs are going down while performance goes up. The next few weeks will determine if this new model truly changes the competitive landscape.
