Kimi K2.5
About
Moonshot Kimi K2.5 available on AWS Bedrock
Capabilities
VisionMultimodalReasoningFunction CallingTool UseStructured OutputsCode Execution
Providers(7)
Compare all →| Provider | Input (per 1M) | Output (per 1M) | Type | |
|---|---|---|---|---|
| Fireworks AI | $0.60 | $3.00 | Serverless | |
| OpenRouter | $0.3827 | $1.72 | Serverless | |
| Together AI | $0.5 | $2.8 | Serverless | |
| Fireworks AI | $0.99 | $4.94 | Serverless | |
| NVIDIA NIM | — | — | Serverless | |
| AWS Bedrock | — | — | Serverless | |
| Replicate API | $0.60 | $3.00 | Serverless |
Benchmark Scores(5)
| Benchmark | Score | Version | Source |
|---|---|---|---|
| Google-Proof Q&A | 87.9 | diamond | Artificial Analysis |
| MMLU PRO | 87.1 | — | https://huggingface.co/spaces/TIGER-Lab/MMLU-Pro |
| Massive Multi-discipline Multimodal Understanding | 84.3 | — | https://huggingface.co/Qwen/Qwen3.5-397B-A17B |
| Instruction-Following Evaluation | 93.9 | — | https://huggingface.co/Qwen/Qwen3.5-397B-A17B |
| BFCL | 68.3 | v4 | https://huggingface.co/Qwen/Qwen3.5-397B-A17B |
Rankings
Specifications
FamilyKimi
Released2026-03-15
Parameters1T (MoE, 384 experts)
Context256K
ArchitectureMixture of Experts
Specializationcoding
Trainingfinetuning
