Head to head
Kimi K3 vs DeepSeek V4 Pro
Which is better, Kimi K3 or DeepSeek V4 Pro?
Kimi K3 finishes ahead of DeepSeek V4 Pro on the index, 85.1 to 83.4. It is the stronger of the two on Craft, while DeepSeek V4 Pro still leads on Control and Value. Both are scored on the same rubric, in the same week, and neither placing is sponsored.
Updated
Kimi K3
20 of 73
85.1
One of the strongest open-weights models, and it reads a million tokens without blinking.
DeepSeek V4 Pro
28 of 73
83.4
DeepSeek's reasoning flagship, now out of preview and outscored by its own Flash model.
The four scores, side by side
| Score | Kimi K3 | DeepSeek V4 Pro | Difference |
|---|---|---|---|
| Craft | 88 | 82 | +6 |
| Speed | 74 | 74 | — |
| Control | 91 | 94 | +3 |
| Value | 78 | 84 | +6 |
| Index score | 85.1 | 83.4 | +1.7 |
Which to pick
Pick Kimi K3 unless Control and Value is what decides it for you, which is exactly where DeepSeek V4 Pro is the better answer.
Kimi K3
Moonshot published K3 at a very large scale, with native vision, an explicit thinking mode and a million-token window, and in October it still ranks among the top open models on both independent leaderboards. Two caveats matter: the licence restricts commercial use above a revenue threshold, so it is open weights rather than open source, and the hosted API costs three dollars in and fifteen out, priced like a Western frontier model rather than a Chinese one.
Where it shines
- Million-token context handled properly rather than nominally
- Native vision and an explicit thinking mode
- Among the highest human-preference ratings of any open model
DeepSeek V4 Pro
V4 Pro left preview in August as V4-Pro-0813: a very large sparse model that thinks before it answers, with MIT-licensed weights and a million-token window. It is still good on hard reasoning, but V4.1 Flash, built on DeepSeek's newer architecture, now scores higher on both independent leaderboards at about a quarter of the price. Pro remains a capable MIT-licensed reasoning model, but for most new work Flash is the better buy.
Where it shines
- MIT licence, including for the reasoning traces
- Very long context, so whole repositories fit in one call
- Off-peak pricing halves the bill