# Kimi K3 — API, capabilities & pricing | Y-API

> Kimi K3 is Moonshot AI’s open-weight multimodal model for coding, knowledge work and long-horizon agents. Its published design emphasizes iterating against repositories, images, logs and runtime feedback rather than generating a single isolated answer.

This is the markdown representation of https://y-api.bestvirtualgoods.com/models/moonshotai/kimi-k3. Generated by `scripts/generate-seo-assets.mjs` from the same copy the page renders — do not edit by hand.

Model ID: `moonshotai/kimi-k3`

Sources reviewed: 2026-10-04. Y-API catalog snapshot: 2026-10-04.

## Choosing this model

Selection advice by Y-API. The checks below are suggested evaluations, not published test results.

### Where to start

Evaluate it in a repository task with a real feedback loop: locate a failure, propose a patch, run tests and revise. Score completed requirements and unintended changes, not the length of the plan.

### What to watch for

An agent harness supplies tools, state and execution limits; the model does not acquire those through this API call alone. Its larger context also makes repeated full-history requests worth budgeting explicitly.

## Model specifications

These are OpenRouter model-level specifications, not a Y-API compatibility test. A particular route may accept fewer input formats, parameters or tokens. Verify the features you need with a small request; a successful text response does not validate vision or tool calling.

| Field | Value |
| --- | --- |
| Context window | 1,048,576 tokens |
| Input | Text, Image, Video |
| Output | Text |
| Listed on OpenRouter | 2026-07-16 |
| Parameters listed by OpenRouter | `frequency_penalty`, `include_reasoning`, `logit_bias`, `logprobs`, `max_tokens`, `min_p`, `presence_penalty`, `reasoning`, `reasoning_effort`, `repetition_penalty`, `response_format`, `seed`, `stop`, `structured_outputs`, `temperature`, `tool_choice`, `tools`, `top_k`, `top_logprobs`, `top_p` |

## Estimate your cost

| Price basis | Input / 1M tokens | Output / 1M tokens |
| --- | --- | --- |
| Account credit | $3.00 | $15.00 |
| Cash equivalent | $0.30 | $1.50 |

$1 paid adds $10 of credit. Cash equivalent = credit consumed ÷ 10; this is not the model publisher’s list price.

Input tokens / request: 1000. Output tokens / request: 500. Number of requests: 1000.

Estimated credit consumed: $10.50. Estimated cash equivalent: $1.05.

Illustrative token budget, not measured usage or a quote. Includes input and output; excludes separate media charges, cache discounts and retries. Include billed reasoning tokens in your output budget where applicable. Actual billing follows usage returned by the service.

### Same token budget, different models

| Model | Estimated credit consumed | Estimated cash equivalent |
| --- | --- | --- |
| [Kimi K3](https://y-api.bestvirtualgoods.com/models/moonshotai/kimi-k3) | $10.50 | $1.05 |
| [Kimi K2.6](https://y-api.bestvirtualgoods.com/models/moonshotai/kimi-k2.6) | $2.95 | $0.295 |
| [Claude Opus 5](https://y-api.bestvirtualgoods.com/models/anthropic/claude-opus-5) | $17.50 | $1.75 |

## Use the API

These minimal, text-only requests use the exact Y-API model ID. They do not demonstrate image, audio, video, file or tool support. The output limit also needs room for reasoning; an empty answer with finish_reason=length can mean the budget was exhausted.

Set the TOKEN environment variable to a key from your console. Run this on your server or locally; never expose a key in browser code or a public repository.

### A task to try

An API times out only on large exports. Design a diagnostic sequence that distinguishes slow SQL, serialization cost and proxy timeouts, without changing production data.

```bash
curl --fail-with-body --silent --show-error --max-time 120 \
  'https://api.y-api.bestvirtualgoods.com/v1/chat/completions' \
  -H "Authorization: Bearer ${TOKEN:?Set TOKEN first}" \
  -H 'Content-Type: application/json' \
  --data-binary @- <<'JSON'
{
  "model": "moonshotai/kimi-k3",
  "messages": [
    {
      "role": "user",
      "content": "An API times out only on large exports. Design a diagnostic sequence that distinguishes slow SQL, serialization cost and proxy timeouts, without changing production data."
    }
  ],
  "max_tokens": 4096
}
JSON
```

```python
import json
import os
import sys
import urllib.error
import urllib.request

payload = {
  "model": "moonshotai/kimi-k3",
  "messages": [
    {
      "role": "user",
      "content": "An API times out only on large exports. Design a diagnostic sequence that distinguishes slow SQL, serialization cost and proxy timeouts, without changing production data."
    }
  ],
  "max_tokens": 4096
}

request = urllib.request.Request(
    "https://api.y-api.bestvirtualgoods.com/v1/chat/completions",
    data=json.dumps(payload).encode("utf-8"),
    headers={
        "Authorization": "Bearer " + os.environ["TOKEN"],
        "Content-Type": "application/json",
    },
    method="POST",
)
try:
    with urllib.request.urlopen(request, timeout=120) as response:
        result = json.load(response)
    print(json.dumps(result, ensure_ascii=False, indent=2))
except urllib.error.HTTPError as error:
    print(error.read().decode("utf-8"), file=sys.stderr)
    raise SystemExit(1)
```

### How to evaluate the result

Require measurements that isolate the three stages and a safe order of investigation. In an agent run, check that each tool result changes the next decision rather than being ignored.

## Before you choose

### Does calling K3 automatically start multiple agents?

No. Multi-agent behavior requires orchestration in your application or client. Set tool permissions, budgets and stopping rules separately from the model selection.

## Sources & scope

Technical facts come from the cited model page and, where available, its linked publisher card. A card describes that checkpoint; it is not proof of the weights a gateway serves. Prices come only from the Y-API catalog. We do not claim measured latency, uptime or benchmark scores for this endpoint.

- [OpenRouter model page & specifications](https://openrouter.ai/moonshotai/kimi-k3)
- [Publisher model card linked by OpenRouter](https://huggingface.co/moonshotai/Kimi-K3)
- [Y-API catalog & prices (JSON)](https://y-api.bestvirtualgoods.com/models.json)

## Models to compare

Compare these alternatives on the same inputs. Their descriptions explain different roles; a lower price does not establish equivalent quality.

### [Kimi K2.6](https://y-api.bestvirtualgoods.com/models/moonshotai/kimi-k2.6)

Kimi K2.6 is a native multimodal model focused on coding, UI generation and agent orchestration. The reference catalog lists text and image inputs, with a smaller context window than the K3 entry.

### [Claude Opus 5](https://y-api.bestvirtualgoods.com/models/anthropic/claude-opus-5)

Claude Opus 5 is positioned by OpenRouter as Anthropic’s model for demanding reasoning, software engineering and extended agent work. The reference entry lists text, image and file input with text output.

## Links

- HTML version of this page: https://y-api.bestvirtualgoods.com/models/moonshotai/kimi-k3
- Site index for agents: https://y-api.bestvirtualgoods.com/llms.txt
- Full reference (single file): https://y-api.bestvirtualgoods.com/llms-full.txt
- OpenAPI 3.1 spec: https://y-api.bestvirtualgoods.com/openapi.json
- Model catalog (JSON, no key needed): https://y-api.bestvirtualgoods.com/models.json
- API base URL: `https://api.y-api.bestvirtualgoods.com/v1`
- Contact: support@bestvirtualgoods.com
- Model catalog: https://y-api.bestvirtualgoods.com/models
- Integration guide: https://y-api.bestvirtualgoods.com/docs