AdvertiseDonate
☰ SectionsCity news, clearly told.
Daily Media Wire · 1 Liberty Square, Suite 301, New Britain, CT

Moonshot AI releases Kimi K3.1 Pro with 33 percent cheaper API calls

The new model scores 55 on the Artificial Analysis Intelligence Index and will join the open-weights field by November 4.

By Madison Morris ·

Illustration for: Moonshot AI releases Kimi K3.1 Pro with 33 percent cheaper API calls
Illustration generated for this article, not a news photograph.

Moonshot AI has launched Kimi K3.1 Pro, a reasoning model that cuts API input costs by 33 percent while scoring 55 on the Artificial Analysis Intelligence Index. This result places the model fourth globally, just three points behind the leading proprietary systems.

The new release builds on the previous Kimi K3 architecture, which features 2.8 trillion total parameters and 104 billion active parameters. Moonshot applied substantial additional post-training compute to the base model to achieve the performance gains. The company scheduled the release of open model weights for November 4, 2026.

If the weight release proceeds as planned, Kimi K3.1 Pro becomes the highest-scoring open-weights model available. That outcome would narrow the performance gap between open and proprietary models from 12 points to 3 points. Currently, only Anthropic's Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full score higher on the index.

Performance gains were most visible in agentic tasks. On the GDPval-AA benchmark, the model reached 1,846 Elo, a 312-point increase over its predecessor. It trails Claude Opus 5.5 at 1,866 Elo and Muse Wonder Full at 1,857 Elo. On Terminal-Bench 4.0, Moonshot reported a score of 65.7 percent, surpassing GPT-6 Astra at 58.2 percent but falling slightly short of Claude Opus 5.5 at 66.4 percent.

Moonshot also reported a reduction in hallucinations. Claim-level hallucinations dropped from 23.4 percent on Kimi K3 to 8.3 percent on the internal Kimi Hallu Bench Hard. The company noted that GPT-6 Astra still leads this specific benchmark at 5.5 percent. In finance, the model gained 7.2 points on Financial Modeling and 11.6 points on Financial Research.

Pricing for the new model sets input tokens at $2 per million and output tokens at $10. Cache hits cost $0.20 per million tokens. Although the model generates 14 percent more output tokens, the cost to run each Intelligence Index task remains $1.98, nearly identical to the $2.00 cost of the previous version.

The architecture includes a one-million-token context window and supports both text and image inputs. Moonshot also added a 3D Make feature on Kimi.com for previewing and exporting models. The new model integrates with Unity, Unreal Engine, and Godot without requiring MCP connectors for Blender and CAD software.

Source: The Decoder

More Tech · Report a business · Got a Story?

Get Daily Media Wire in your inboxCity news, clearly told. Free, a few times a week, unsubscribe any time.

Related stories

Get Daily Media Wire in your inboxCity news, clearly told. Free, a few times a week, unsubscribe any time.