Kimi K3
Kimi K3The open frontier model. 2.8T parameters, 1M-token context, top benchmarks scores.
Model details
Kimi K3 is Moonshot AI’s flagship 2.8-trillion-parameter Mixture-of-Experts (MoE) large language model. It is built specifically to handle complex reasoning, long-horizon coding, and agentic knowledge work.
Key specifications and capabilities include:
1M Token Context Window: Supports massive context lengths with native vision capabilities, enabling processing of entire codebases, research papers, and multi-modal inputs.
Long-Horizon Coding & Engineering: Built to run extended coding sessions autonomously with minimal supervision.
Native Multimodality: Combines software logic with visual reasoning, allowing it to leverage screenshots for tasks in game design, frontend UI, CAD, and dashboard creation.
Reasoning Modes: Operates with native thinking capabilities, allowing users to configure reasoning effort levels (
low,high, ormax) depending on task complexity.
1from openai import OpenAI
2
3client = OpenAI(
4 api_key="YOUR_API_KEY",
5 base_url="https://inference.baseten.co/v1"
6)
7
8response = client.chat.completions.create(
9 model="moonshotai/Kimi-K3",
10 messages=[
11 {
12 "role":"user",
13 "content":[
14 {"text":"Describe this image in one sentence.","type":"text"},
15 {"type":"image_url",
16 "image_url":{"url":"https://picsum.photos/id/237/200/300"}
17 }
18 ]
19 }
20 ],
21)
22
23print(response.to_json())
241{
2 "id": "chatcmpl-xxxx",
3 "choices": [
4 {
5 "finish_reason": "stop",
6 "index": 0,
7 "logprobs": "None",
8 "message": {
9 "content": "<CONTENT_TEXT>",
10 "refusal": "None",
11 "role": "assistant",
12 "annotations": "None",
13 "audio": "None",
14 "function_call": "None",
15 "reasoning_content": "<REASONING_TEXT>"
16 },
17 "stop_reason": 163586,
18 "token_ids": "None",
19 "routed_experts": "None"
20 }
21 ],
22 "model": "moonshotai/Kimi-K3",
23 "object": "chat.completion",
24 "usage": {
25 "completion_tokens": 82,
26 "prompt_tokens": 189,
27 "total_tokens": 271,
28 "prompt_tokens_details": {
29 "cached_tokens": 0,
30 "created_cache_tokens": 0,
31 "multimodal_tokens": {
32 "image": 96
33 }
34 }
35 },
36 "prompt_logprobs": "None",
37 "prompt_token_ids": "None",
38 "prompt_text": "None",
39 "kv_transfer_params": "None",
40 "ec_transfer_params": "None",
41 "metrics": "None"
42}
