# Deepseek V3.1 NVFP4 vs Kimi K2.7 Code

On provider list prices, Deepseek V3.1 NVFP4 costs $0.60 per million input tokens against $0.95 for Kimi K2.7 Code: 1.6x apart. Output is $1.70 against $4 (2.4x). On Allocate both bill at list plus the 7% transaction fee.

## Specifications

| | Deepseek V3.1 NVFP4 | Kimi K2.7 Code |
| --- | --- | --- |
| Lab | DeepSeek | Moonshot AI |
| Access | Open weights | Open weights |
| Context window | 128K tokens | 256K tokens |
| List price, input | $0.60 / M tokens | $0.95 / M tokens |
| List price, output | $1.70 / M tokens | $4 / M tokens |
| Cached input | n/a | $0.19 / M tokens |
| License | MIT | Not listed |
| Fine-tunable | Yes | Yes |

Specifications and provider list prices from the Allocate catalog, checked 2026-07-08. Billed price is list plus the 7% transaction fee.

## What the numbers say

Take 1,000,000 requests a month at 1,200 input and 350 output tokens each. That workload costs $1,315 a month on Deepseek V3.1 NVFP4 and $2,540 on Kimi K2.7 Code at list: a gap of $1,225, or 1.9x.

Kimi K2.7 Code reads 256K tokens per request against 128K for Deepseek V3.1 NVFP4, 2.0x the window. That decides which one can take whole documents without splitting them.

## Choose Deepseek V3.1 NVFP4 for

- The lower list price ($0.60 in / $1.70 out per M tokens)
- Fine-tuning under a permissive license (MIT)

## Choose Kimi K2.7 Code for

- The longer context window (256K vs 128K tokens)
- Published cached-input pricing ($0.19 per M tokens)

## Common questions

### Which is cheaper, Deepseek V3.1 NVFP4 or Kimi K2.7 Code?

Deepseek V3.1 NVFP4, on this workload shape. At list prices it is $0.60/$1.70 per million tokens in and out against $0.95/$4 for Kimi K2.7 Code. Billed on Allocate: $0.64/$1.82 against $1.02/$4.28, list plus 7%.

### Which has the bigger context window?

Kimi K2.7 Code: 262,144 tokens (256K) against 131,072 (128K) for Deepseek V3.1 NVFP4.

### Can I fine-tune Deepseek V3.1 NVFP4 or Kimi K2.7 Code?

Both publish open weights (Deepseek V3.1 NVFP4: MIT; Kimi K2.7 Code: Not listed), so both can be fine-tuned. On Allocate the trained weights stay inside your boundary and belong to you.

---

[HTML page](https://allocate.network/compare/deepseek-deepseek-v3-1-vs-moonshotai-kimi-k2-7-code) · [Deepseek V3.1 NVFP4](https://allocate.network/models/deepseek-deepseek-v3-1.md) · [Kimi K2.7 Code](https://allocate.network/models/moonshotai-kimi-k2-7-code.md) · [Machine-readable catalog](https://allocate.network/catalog.json)