Kimi K3.1 Pro vs Kimi K3: API pricing, caching, and hosts
Pro costs less than K3 on the half-cached example request. Pro's open weights also have a third party host listed at $0.3287 input and $8.115 output. Read the custom license before commercial use.
Kimi K3.1 Pro lists 33% less than Kimi K3 for input and 33% less for output. Pro is post-trained from K3 and sits in the same 2.8T class, so a move is a question of cost and testing, not a new vendor and a new integration. The price table is the easy half. The harder half is whether Pro finishes your tasks in fewer tokens, or with fewer retries, than K3 does. This page prices both models on three workloads and lists what else changes when you switch.
Compare Pro and K3 on base rates, cache reads, and where open weights can run.
By TechCompare · Updated
MoonshotAI: Kimi K3.1 Pro
Moonshot AI's own price
- Input
- $2
- Output
- $10
- Cached
- $0.20
per 1M tokens
Price checked 2026-11-07
MoonshotAI: Kimi K3
Moonshot AI's own price
- Input
- $3
- Output
- $15
- Cached
- $0.30
per 1M tokens
Price checked 2026-11-07
Cost Comparison
Based on 100,000 input tokens (50% cached), 5,000 output tokens, and 100 requests.Prices come live from OpenRouter at each maker's own price where it's listed. Temporary deals are left out.
Side-by-side specs
| Spec | Kimi K3.1 Pro | Kimi K3 |
|---|---|---|
| Standard input per 1M tokens | $2.00 (better on this spec) | $3.00 |
| Standard output per 1M tokens | $10.00 (better on this spec) | $15.00 |
| Cache reads per 1M tokens | $0.20 (better on this spec) | $0.30 |
| 100K input + 5K output, no cache hits | $0.25 (better on this spec) | $0.375 |
| 100K input + 5K output, half the input cached | $0.16 (better on this spec) | $0.24 |
| Agent session, 30 turns, mostly cached | $2.268 (better on this spec) | $3.402 |
| Batch input / output per 1M | None | None |
| Third party host, input / output per 1M | $0.3287 / $8.115 | $0.1634 / $10.1489 |
| Context window | 1,048,576 tokens | Check Moonshot's docs |
| Weights | Open, custom license | Check Moonshot's docs |
How they differ
Kimi K3.1 Pro charges $2 per million input tokens, $10 per million output tokens and $0.20 for cache reads. Kimi K3 charges $3, $15 and $0.30. One request with 100,000 input tokens and 5,000 output tokens costs $0.25 on Pro and $0.375 on K3 when nothing is cached. With half the input read from cache, it costs $0.16 and $0.24. Over 10,000 of those half-cached requests, moving from K3 to Pro saves $800. Three workloads show where the gap matters. An agent session of 30 turns, each sending 120,000 input tokens with nearly all of it cached and producing 3,000 output tokens, costs $2.268 on Pro and $3.402 on K3. A document pipeline that processes 10,000 files at 30,000 input tokens and 2,000 output tokens each, with no cache hits, costs $800 and $1,200. A chat product handling 100,000 conversations at 8,000 input tokens and 800 output tokens, with about two thirds of the input cached, costs $1,464 and $2,196. Neither lists batch prices right now, so the pipeline job pays the standard rates. These examples leave out retries, tool fees and the reasoning tokens a hard task adds to the output side. The embedded calculator uses live OpenRouter rates, which can depend on provider routing.
Verdict
Treat Pro as a drop in candidate and test it on the tasks where K3 struggles or costs you retries. On price alone, Pro has the edge at every cache share, so the switch pays for itself unless it needs more attempts per accepted result. Compare billed tokens for each accepted result, not the price per million. A model that thinks longer on the same task can erase a lower rate. Pro adds sparse attention on its latent layers and ships under a custom license, so check both before you commit production traffic. The license terms aren't summarized here.
Which should you pick?
Choose Kimi K3.1 Pro
Move to Pro when your tests show it finishes the same tasks with no more tokens or retries than K3. It accepts text, image and video input, supports tools and reasoning, and offers a 1,048,576 token context. Its weights are open, and a third party host lists it at $0.3287 input and $8.115 output with fp4 weights, so confirm that host's quantization on your own tasks. Teams that may self host later should note that the memory estimates for Pro and K3 sit close together.
Moonshot's pricing pageChoose Kimi K3
Stay on K3 when it already meets your quality bar and your evaluation suite is tuned to it. A price gap rarely repays the work of rerunning evaluations, retuning prompts and reading Pro's license. A third party host lists K3 at $0.1634 input and $10.1489 output right now, so look at that before you move. Check Moonshot's documentation for K3's context limit before sending very long prompts, since this page doesn't state it.