Model comparison
Kimi K2 vs DeepSeek-V3
- Input
- $0.55 / 1M tokens
- Output
- $2.20 / 1M tokens
- Context
- 128K
- Released
- 2025-07-11
- Input
- $0.27 / 1M tokens (cache miss)
- Output
- $1.10 / 1M tokens
- Context
- 128K tokens
- Released
- 2024-12-26
Kimi K2 scores higher on 6 of the 6 benchmarks both models report; DeepSeek-V3 is cheaper on input tokens ($0.3 vs $0.6 per 1M, 2×).
Specs and pricing side by side
| Spec | Kimi K2 | DeepSeek-V3 |
|---|---|---|
| Maker | Moonshot AI (Kimi) | DeepSeek |
| Released | 2025-07-11 | 2024-12-26 |
| Context window | 128K | 128K tokens |
| Max output | 128K | not published |
| Input price | $0.55 / 1M tokens | $0.27 / 1M tokens (cache miss) |
| Output price | $2.20 / 1M tokens | $1.10 / 1M tokens |
| Parameters | 1T total / 32B active (MoE) | 671B total / 37B active per token (Mixture-of-Experts) |
| License | Modified MIT License | DeepSeek Model License (commercial use permitted); code under MIT |
| Weights | Open weights | Open weightssame |
| Modalities | Text | text |
Not published by either maker: knowledge cutoff.
Benchmarks both models report
| Benchmark | Kimi K2 | DeepSeek-V3 |
|---|---|---|
| SWE-bench Verified | 65.8% | 42% |
| AIME 2024 | 69.6% | 39.2% |
| MATH-500 | 97.4% | 90.2% |
| MMLU | 89.5% | 88.5% |
| MMLU-Pro | 81.1% | 75.9% |
| GPQA-Diamond | 75.1% | 59.1% |
Scores are as published by each maker or its cited source; harness, reasoning effort and date can differ between rows, so treat small gaps as ties. Sources are listed on each model page.
Which one should you pick?
Pick Kimi K2 if…
- You care about SWE-bench Verified, AIME 2024, MATH-500 (it scores higher on 6 shared benchmarks).
- You need Text input, which DeepSeek-V3 lacks.
- Your use case is: agentic coding assistants and automated software-engineering workflows.
Pick DeepSeek-V3 if…
- Cost matters: input tokens cost 2× less ($0.3 vs $0.6 per 1M).
- You need text input, which Kimi K2 lacks.
- Your use case is: general-purpose chat and instruction following at low cost.
FAQ
Which is better, Kimi K2 or DeepSeek-V3?
Kimi K2 scores higher on 6 of the 6 benchmarks both models report; DeepSeek-V3 is cheaper on input tokens ($0.3 vs $0.6 per 1M, 2×). The better pick depends on whether you value benchmark scores, price per token, context length or openness for your workload.
How much cheaper is DeepSeek-V3?
Kimi K2: $0.55 / 1M tokens input, $2.20 / 1M tokens output. DeepSeek-V3: $0.27 / 1M tokens (cache miss) input, $1.10 / 1M tokens output (list API prices as published by each maker).
What are the context windows of Kimi K2 and DeepSeek-V3?
Kimi K2 offers 128K with up to 128K output; DeepSeek-V3 offers 128K tokens.
Is Kimi K2 newer than DeepSeek-V3?
Kimi K2 was released 2025-07-11; DeepSeek-V3 was released 2024-12-26.