Kimi K3
- Input / 1M tokens
- $0.19
- Output / 1M tokens
- $16.00
- Providers
- 18
- Uptime, 24 h
- 100%
API models · Side by side
Kimi K3 vs Grok 4.7
Kimi K3 and Grok 4.7 charge for very different parts of a request. Kimi has the lower input rate, while Grok has the lower output rate on ordinary prompts. A long reference document with a short answer can favour Kimi; a brief instruction that produces pages of text can favour Grok.
Grok 4.7
- Input / 1M tokens
- $2.00
- Output / 1M tokens
- $6.00
- Providers
- 2
- Uptime, 24 h
- 100%
The short answer
Choose Kimi when direct video input, more context or a future self-hosted deployment matters. Consider Grok for output-heavy text work and applications already using its developer tools. For cost alone, compare complete requests: Grok’s higher long-prompt tier can reverse the advantage of its cheaper output tokens.
Estimate your monthly bill
Enter your own usage. We apply each model's current rates, including long-prompt pricing, for a 30-day month.
Try:
Kimi K3$251.52per month
Grok 4.7$210.00per month
A token is roughly three-quarters of an English word. Cached input, reasoning tokens, tools and media are not included.
Compare the details
Compare context, supported inputs and usage costs. Highlighted rows show a difference.
| Feature | Kimi K3 | Grok 4.7 |
|---|---|---|
| Input formats | image, text, video | file, image, text |
| Output formats | text | text |
| Context / position limit | 1,048,576 | 500,000 |
| Endpoint output limit (tokens) | 943,718 | 450,000 |
| Developer features | frequency penalty, Reasoning output, logit bias, logprobs, Output limit, min p, presence penalty, Reasoning controls, Thinking effort, repetition penalty, Response format, Seed, Stop sequences, Structured output, Temperature, Tool selection, Tool calling, top k, top logprobs, Sampling controls | Reasoning output, logprobs, Output limit, Reasoning controls, Thinking effort, Response format, Seed, Stop sequences, Structured output, Temperature, Tool selection, Tool calling, top k, top logprobs, Sampling controls |
| Cached input / 1M tokens | $0.18 | $0.50 |
| USD input / 1M tokens | $0.19 | $2.00 |
| USD output / 1M tokens | $16.00 | $6.00 |
| Available API providers | 18 | 2 |
| Best provider uptime, last 24 h | 100% | 100% |
| Price range across providers (per 1M tokens) | Input $0.18–$4.50, output $9.50–$22.50 across 18 providers | Set by the developer |
| Model identifier | moonshotai/kimi-k3 | x-ai/grok-4.7 |
API pricing
Prices are in USD per million tokens. Input is what you send; output is what the model generates. Chat subscriptions and API usage are billed separately.
Grok 4.7: long-prompt rates
- Long-prompt threshold: 200,000 prompt tokens. Input: $4.00 / 1M. Output: $12.00 / 1M.
Tools, images and additional reasoning can add charges. The examples below cover input and output tokens only.
Compare local modelsWhat those prices mean in practice
Each example uses the same token budget for both models. These are calculations, not measured workloads. Cache discounts, media, tools and extra reasoning are excluded.
| Workload | Kimi K3 | Grok 4.7 |
|---|---|---|
| 1,000 short requestsPer request: 2,000 input + 500 output tokens | $8.38 | $7.00 |
| 100 document summariesPer request: 20,000 input + 1,000 output tokens | $1.98 | $4.60 |
| 10 long-document requestsPer request: 300,000 input + 2,000 output tokens | $0.90 | $12.24 |
Estimate = requests × (input tokens × input price + output tokens × output price) ÷ 1,000,000. Long-prompt rates are applied where the pricing data provides them. Real requests can use different amounts of output.
Beyond the price tag
Cheap input and cheap output are different advantages
At the compared API rates, Kimi costs $0.152 per million input tokens and $20 per million output tokens. Grok costs $2 and $6 before its long-prompt tier. For 1,000 requests with 2,000 input tokens and 500 output tokens each, that is about $10.30 for Kimi and $7 for Grok. Grok is already cheaper for that mix of reading and writing, and the gap widens as answers get longer.
Long answers can make Grok less expensive
Keep the input at 2,000 tokens but ask for 2,000 output tokens per request. Across 1,000 requests, Kimi costs about $40.30 and Grok $16 at the base rates. That makes output length a useful budget control for drafting and code generation. A shorter instruction does not necessarily mean a cheap task if the model then writes a long answer. These are equal-token calculations, not measurements of how much either model would actually generate.
Document extraction can point the other way
For 100 requests with 20,000 input tokens and 1,000 output tokens each, Kimi costs about $2.30 versus $4.60 for Grok. That mix is closer to reading a substantial document and returning a concise result. The same source file may produce different token counts on the two models, so use each service’s usage figures when budgeting a real application. Include retries and rejected answers before deciding what an accepted result costs.
Grok’s long-prompt tier changes the calculation
Grok’s rate increases to $4 per million input tokens and $12 per million output tokens in its long-prompt tier, which starts at 200,000 prompt tokens in the compared pricing data. A request with 300,000 input tokens and 2,000 output tokens therefore costs $1.224 on Grok. Kimi has no prompt-length surcharge in the compared rate, making the same token budget about $0.086. Both totals exclude caching, tools and additional reasoning. The calculator applies the available tiers automatically.
Video support and context make a practical difference
Kimi accepts text, images and video. The Grok route compared here accepts text, images and files, without direct video input. For a recorded interface problem, Kimi can take the recording; Grok needs another preparation step, such as selected frames and a description. Kimi’s context allowance is 1,048,576 tokens versus Grok’s 500,000. That extra space can accommodate a larger reference set, but it does not establish that an answer finds every important detail.
Reasoning and tools belong in the integration plan
Kimi always enables thinking, with low, high and max effort settings and max as the default. Grok supports low, medium, high and xhigh, defaulting to high. Do not assume the defaults spend the same amount of work. Grok’s developer API also offers tools such as web search and code execution; a model request through another service may expose a different tool set. Price those operations separately from the text-token examples and confirm that your chosen route supports the features your application uses.
Kimi offers an option to operate the model yourself
Kimi’s weights are available under the Kimi K3 License. That gives teams a deployment option beyond buying API calls, with responsibility for hardware, serving software and operation. Its 2.8 trillion total parameters are distinct from the 104 billion active parameters: the active count is not the amount of model data a server needs to hold. Grok is a hosted choice in this comparison. If operating the model is a requirement, assess the exact Kimi files and runtime rather than turning an API price into a GPU budget.
Choose a model for the work you finish
A screenshot triage feature can check whether the answer identifies the error and gives an appropriate next step. A drafting feature can count the editing required before publication. A coding workflow can check whether the proposed change meets its requirements. Those are different reasons to accept or reject an answer. Keep that standard consistent across both models, and compare complete task costs rather than treating a context size or price label as a quality score.
Common questions
Which costs less for a short chatbot reply?
It depends on the token mix. At the base rates used here, 2,000 input tokens and 500 output tokens cost less on Grok ($7 against $10.30 per 1,000 requests), and longer answers widen the gap. Kimi is cheaper when the request is mostly reading, such as a long document with a short answer. Reasoning, retries and tools can change the actual bill.
Can both models produce video?
No. Both compared endpoints return text. Kimi’s video support means it can read video as input; that is different from generating a video. Grok’s separate media-generation products are outside this comparison.
Can either model use the whole context as its answer?
A context allowance is not a promised response length. The table shows the output limits exposed by the compared hosting endpoints. Input, output, reasoning settings and request constraints still need to fit the service you use.
Are these Kimi or Grok app subscription prices?
No. These are API token charges for applications. Chat subscriptions have separate prices and usage rules, and do not replace an API budget.