Kimi K2 Base: pricing, context window and benchmarks
Kimi K2 Base is a multimodal model from Moonshot AI. It offers a 65,536-token context window, costs $4.82 per million input tokens and $16.90 per million output tokens, and averages 79.2 across the index's benchmark suite.
Pricing
Kimi K2 Base costs $4.82 per million input tokens and $16.90 per million output tokens. A blended 3:1 input-to-output mix works out to about $7.84 per million tokens. Cached input is listed at $0.57 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $4.82 | per 1M tokens |
| Output | $16.90 | per 1M tokens |
| Cached input | $0.57 | per 1M tokens |
| Blended 3:1 | $7.84 | estimate per 1M tokens |
| 1M in + 1M out | $21.72 | estimate |
Benchmarks
Scores below are modelled estimates on a 0–100 scale. Use them to compare relative position, not as official vendor results.
| Benchmark | Score | Rank | Percentile | Relative |
|---|---|---|---|---|
| MMLU-Pro | 81.8 | 126 | 79th | |
| GPQA Diamond | 76.1 | 120 | 80th | |
| SWE-bench Verified | 70.2 | 106 | 82th | |
| HumanEval | 84.9 | 121 | 80th | |
| MATH-500 | 80.8 | 194 | 68th | |
| AIME 2025 | 75.3 | 149 | 75th | |
| LiveCodeBench | 69.9 | 192 | 68th | |
| IFEval | 86.8 | 139 | 77th | |
| TruthfulQA | 83.2 | 104 | 83th | |
| HellaSwag | 88.5 | 124 | 79th | |
| MMMU | 73.5 | 204 | 66th | |
| Arena Elo | 1432.3 | 116 | 81th |
What Kimi K2 Base is good for
Kimi K2 Base is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $4.82 per million input tokens, it sits in the mid-range of the index, and the 65,536-token context window is enough for large files and multi-file diffs.
Strengths
- Open weights make self-hosting, fine-tuning and data-residency review possible.
- Accepts more than text, including image input where the provider supports it.
- Coding benchmarks are a relative strength within its tier.
Trade-offs
- No standout weakness in the indexed specifications, but real-world behaviour still needs testing.
How it compares
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Llama 4 Scout | Meta AI | 200000 | $4.90 | 79.3 |
| Reka Core Max | Reka AI | 128000 | $13.41 | 79.3 |
| Claude Haiku 4.5 Vision | Anthropic | 128000 | $13.01 | 79.2 |
| Kimi K2 Base | Moonshot AI | 65536 | $4.82 | 79.2 |
| Llama 4 Scout Instruct | Meta AI | 32768 | $4.59 | 79.1 |
| Sonar Pro | Perplexity | 128000 | $13.01 | 79.1 |
FAQ
How much does Kimi K2 Base cost?
Kimi K2 Base is listed at $4.82 per million input tokens and $16.90 per million output tokens. Cached input is $0.57 per million tokens. A blended 3:1 input-to-output mix is about $7.84 per million tokens.
What is the context window of Kimi K2 Base?
Kimi K2 Base has a context window of 65,536 tokens and can emit up to 32,768 output tokens per request in this index.
Is Kimi K2 Base open weights?
Yes. Kimi K2 Base is listed as an open-weight model under a open weights licence.
How does Kimi K2 Base score on benchmarks?
Kimi K2 Base averages 79.2 across the scored benchmarks in this index, ranking #143 of 600 models. Its strongest relative result is Arena Elo at 1432.3.
What is Kimi K2 Base best for?
Kimi K2 Base is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $4.82 per million input tokens, it sits in the mid-range of the index, and the 65,536-token context window is enough for large files and multi-file diffs.