Kimi K2 0905 Preview is an advanced update of the Kimi K2 model, engineered for high-performance in intelligent agent creation, multi-turn conversational AI, and complex analytical tasks. This version extends the context window to 262,144 tokens and integrates enhanced request caching, delivering unprecedented efficiency and depth in natural language understanding and reasoning. It is tailored for corporate assistants, agent-based workflows, and advanced reasoning applications requiring extensive context and memory.
Across five distinct evaluations, including SWE-bench Verified, Multilingual, and SWE-Dev, it achieves higher average scores than both Kimi K2-0711 and Claude Sonnet 4. Each score represents the average of five rigorous test runs, ensuring statistical reliability.
.webp)
vs GPT-4 Turbo: Kimi-K2-0905 offers double the context length (262K vs. 128K) and superior caching for repetitive enterprise queries. While GPT-4 excels in general creativity, Kimi-K2-0905 is optimized for structured reasoning and agent reliability.
vs Claude 3.5 Sonnet: Both deliver strong analytical performance, but Kimi-K2-0905 provides faster inference on long contexts and native support for stateful agent memory. Claude favors conversational fluency; Kimi prioritizes task completion.
vs Llama 3 70B: Llama 3 is ideal for customization, but lacks built-in long-context optimization and enterprise tooling. Kimi-K2-0905 delivers out-of-the-box performance with managed infrastructure, caching, and compliance.
vs Gemini 1.5 Pro: Gemini matches Kimi in context length, but Kimi-K2-0905 shows lower latency in cached scenarios and better tool-integration for agentic loops. Gemini leads in multimodal tasks; Kimi dominates in text-centric enterprise reasoning.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "moonshotai/kimi-k2-0905", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "moonshotai/kimi-k2-0905", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"moonshotai/kimi-k2-0905","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Terminal-Bench | 44.5% | Autonomous shell/terminal task completion | Source | July 13, 2026 |
| SWE-bench Verified | 69.2% | Resolving verified real GitHub issues | Source | July 13, 2026 |
| Intelligence | 15.3 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Math | 57.3 | Composite score across standardised mathematics evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Kimi K2 0905 Preview This page | Reasoning + agents | |||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents |
Kimi K2 0905 Preview has a 262,144 tokens context window and can return up to 100,352 tokens.
Kimi K2 0905 Preview takes image, text as input and returns text.
Use moonshotai/kimi-k2-0905 as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
Kimi K2 0905 Preview is priced at input $1.3754 / 1M tokens, output $4.1262 / 1M tokens.
Yes, Kimi K2 0905 Preview can stream responses as they are generated.
Yes, Kimi K2 0905 Preview accepts image input alongside text.
Kimi K2 0905 Preview was built by Moonshot AI.
Yes, it supports tool calling, including parallel tool calls and function calling.
