Moonshot AI
Moonshot AI’s premier flagship model with 2.8 trillion parameters, Kimi Delta Attention, 1M context, and open weights.
1.049M ctx$3.00 / 1M tok (2.8T Open Weights)
Introducing Kimi K2.6Moonshot AIReleased April 15, 2026
Kimi K2.6
General-purpose model supporting text, image, and video inputs, switchable thinking modes, and autonomous agent tasks over 256k context.
reasoning-llmProprietary API$0.95 / 1M tok (Vision & Agentic)Context: 262.144K (262,144 tokens)Arena ELO: 1445
Technical Specifications
Architecture Type
General-Purpose Multimodal Agent Transformer
Total Parameters
Frontier Scale
Context Window
262.144K (262,144 tokens)
Max Output Tokens
65.536K (65,536 tokens)
Knowledge Cutoff
March 2026
Supported Modalities
text, code, vision, reasoning
License & Access
Moonshot AI Terms of Service
Benchmark Evaluations
Mmbench
87.5
Mmlu Pro
84.8
Humaneval
94.6
Swe Bench Verified
68.4
Chatbot Arena ELO
1445
Deep Architectural Overview
Kimi K2.6 possesses stronger, more stable long-term code writing capabilities, significantly improved instruction compliance, and autonomous self-correction across complex software engineering projects in Rust, Go, Python, and TypeScript.
Strengths & Considerations
Core Strengths
- Multimodal support for text, images, and video sequences
- Switchable thinking and non-thinking modes
- 256k context window and 64k output buffer
- 68.4% SWE-bench Verified coding capability
Known Limitations
- Succeeded by Kimi K2.7 Code for pure programming tasks and K3 for 1M context
Token & API Pricing
Input Tokens (1M)$0.95
Output Tokens (1M)$4.00
Cached Input (1M)$0.1600
Pricing is verified directly against Moonshot AI's developer documentation and API rate sheets.
Similar & Alternative Models
Explore other frontier models from Moonshot AI and comparable reasoning engines.
Moonshot AI
High-speed variant of Kimi K2.7 Code with generation speeds of 180 to 260 tokens per second for instantaneous coding feedback.
262.144K ctx$1.90 / 1M tok (180-260 tok/s)
Moonshot AI
Dedicated coding model delivering higher task success rates, tighter instruction following, and a 30% reduction in overthinking.
262.144K ctx$0.95 / 1M tok (Dedicated Coding)