Kimi K3.1 Pro
Kimi K3.1 Pro is Kimi's 3T-class open-weight flagship with a 1M-token context, built for long-horizon coding, 3D modeling in Blender and CAD, and finance.
Model Overview
Capabilities, design details, and architectural traits
Kimi K3.1 Pro - 3T-class open-weight model for coding, 3D creation and finance
Kimi K3.1 Pro is the first model of the K3.1 series and Kimi's most capable 3T-class open-weight model, with a 1-million-token context window. It is a post-trained version of Kimi K3, built with large-scale joint RL across hundreds of downstream domains. Its defining focus is long-horizon agentic work: sustained coding sessions, 3D creation, and end-to-end financial workflows.
| Defining trait | What it means |
|---|---|
| 3D creation without MCP connectors | Writes and runs scripts directly in Blender, SolidWorks, Rhino and other professional modeling tools, inferring shape, proportions and spatial relationships from text, sketches or reference images. Strong at parametric modeling such as mechanical parts, architecture and configurable products. |
| Game engine coding end to end | Works fluently across Unity, Unreal Engine and Godot, combining them with web search and 3D generation to build complete playable experiences, from terrain and boss fights to full project builds. |
| End-to-end financial workflows | Delivers complete finance work: retrieval over real financial documents, cross-document verification, multi-step numerical reasoning, spreadsheet modeling and auditable investment research. |
| Always-on thinking with adjustable effort | Thinking mode is always enabled, with a top-level reasoning_effort setting offering low, high and max (default max). |
| Highspeed twin variant | Kimi K3.1 Pro Highspeed delivers the same intelligence with roughly 2-3x faster output. |
| Open weights with safety controls | Full model weights are released openly, with offensive-security data removed from mid-training and RL to keep attack capability below what the training pipeline would otherwise produce. |
Safety-controlled open-weight release
Before publishing the weights, the team traced how offensive-security capability emerged from general training data, then identified and removed cyber-offense-related data from mid-training and RL while keeping general programming tasks intact. The stated principle is to not push the frontier of offensive security in open weights, while letting defensive teams run the model on infrastructure they control.
API surface
The model accepts image and video input, supports strict JSON-schema structured output, and offers a Partial Mode that continues generation from a supplied text prefix. Streaming separates reasoning deltas from final-answer content.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Kimi | Anthropic | Anthropic |
| Release Date | October 23, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 1.0M Best Context Window | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $2 Best Input Pricing | $4 | $2 Best Input Pricing |
| Output Pricing | $10 Best Output Pricing | $20 | $10 Best Output Pricing |
| Modalities | |||
| Inputs | textimagevideo | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 55.1 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Kimi
Other models by Kimi
Top AI Models
Leading alternatives by intelligence score
