Kimi K2 0905. Kimi-k2 is a Mixture-of-Experts (MoE) foundation model with exceptional coding and agent capabilities, featuring 1 trillion total parameters and 32 billion activated parameters. In benchmark evaluations covering general knowledge reasoning, programming, mathematics, and agent-related tasks, the K2 model outperforms other leading open-source models. Quantized at FP8.
Added Sep 25, 2025
Model weightsContext Window
256.0K
Max Output
262.1K
Avg output tokens (7d)
444 tokens
Input Price (Auto)
$0.40/1M
Output Price (Auto)
$1.80/1M
Cache Read (Auto)
$0.20/1M
Capabilities
Benchmarks
Benchmarks
Performance metrics and benchmarks
Sourced from Artificial Analysis.
Intelligence Index
15.3
Agentic work
T²-Bench Telecom (legacy)
Legacy fallback · Conversational AI agents in dual-control scenarios
73.4%
Better than 65% of models compared
Document reasoning
AA-LCR v1.1
Long context reasoning with updated grading
53.0%
Better than 49% of models compared
Reasoning
HLE
Humanity's Last Exam
6.4%
Better than 42% of models compared
IFBench
Instruction-following benchmark
41.7%
Better than 43% of models compared
Coding
Terminal-Bench Hard (legacy)
Legacy fallback · Agentic coding and terminal use
23.5%
Better than 63% of models compared
LiveCodeBench
Contamination-free coding benchmark
61.0%
Better than 67% of models compared
Math
AIME 2025
American Invitational Mathematics Examination 2025
57.3%
Better than 54% of models compared
Knowledge
MMLU-Pro
Professional and academic subject knowledge
81.9%
Better than 77% of models compared
Legacy benchmarks
GPQA Diamond (legacy)
Graduate-level scientific reasoning
76.7%
Better than 60% of models compared
AA-LCR (unversioned / legacy)
Long context reasoning evaluation
53.0%
Better than 49% of models compared
Last updated Sep 13, 2026
Artificial AnalysisProviders
Auto routing is available for this model. Explicit provider selection is not available.
Loading provider options…
Related text models
Compare Kimi K2 0905 with similar models from the same provider or model family.
Kimi K2 Instruct
moonshotai/kimi-k2-instructKimi K2 0905. Kimi-k2 is a Mixture-of-Experts (MoE) foundation model with exceptional coding and agent capabilities, featuring 1 trillion total parameters and 32 billion activated parameters. In benchmark evaluations covering general knowledge reasoning, programming, mathematics, and agent-related tasks, the K2 model outperforms other leading open-source models. Quantized at FP8.
Kimi K2 0711
moonshotai/kimi-k2-instruct-0711Kimi K2 0711 version. Kimi-k2 is a Mixture-of-Experts (MoE) foundation model with exceptional coding and agent capabilities, featuring 1 trillion total parameters and 32 billion activated parameters. In benchmark evaluations covering general knowledge reasoning, programming, mathematics, and agent-related tasks, the K2 model outperforms other leading open-source models. Quantized at FP8.
Kimi K3
moonshotai/kimi-k3Kimi K3 is Moonshot AI's open-weight, always-thinking multimodal model for long-context reasoning, coding, tool use, and native image/video understanding.
Kimi K2.7 Code High-Speed
moonshotai/kimi-k2.7-code-highspeedKimi K2.7 Code High-Speed is the accelerated coding-focused variant tuned for fast agentic software engineering. It targets roughly 180 tokens per second, with short-context responses reaching up to about 260 tokens per second for rapid coding iterations.
Kimi K2.7 Code
moonshotai/kimi-k2.7-codeKimi K2.7 Code is Moonshot AI's coding-focused agentic model built for long-horizon software engineering workflows. It supports native image input, tool calling, and forced thinking mode; instant/non-thinking mode is not supported.
Kimi Latest
moonshotai/kimi-latestCompatibility alias that routes to the newest Kimi model. Currently routes to Kimi K3.