Kimi K2
Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass.
kimi-k2Released 11 Jul 2025 (Jul 2025)Pricing
List prices at Vercel AI Gateway, in US dollars per million tokens.
| Tier | Input / 1M | Output / 1M | |
|---|---|---|---|
| Standard | $0.55 | $2.20 |
What it costs a month
At Vercel AI Gateway list prices, 30 days.
- Chat assistant5,000 requests a day, 2K in / 0.5K out$330
- RAG over documents2,000 requests a day, 8K in / 0.6K out$343
- Coding agent1,000 requests a day, 40K in / 2K out$792
- Overnight batch20,000 requests a day, 4K in / 0.4K out$1,848
Where to run it
2 hosts serve Kimi K2. The comparison is on blended price, 3 parts input to 1 part output.
| Host | In / out per 1M | Model id | |
|---|---|---|---|
Vercel AI GatewayGateway kimi-k2 | $0.55 / $2.20 | kimi-k2 | |
| $0.60 / $2.50 +12% vs headline | kimi-k2-0905 |
Capabilities
Core
- Streaming
- Function calling
- JSON mode
- Structured outputs
Multimodal
- Vision
- Image input
- Audio
- Image generation
- Video
Advanced
- Reasoning
- Tool calling
- MCP
- Prompt caching
- Batch API
- Embeddings
- Takes in
- Text
- Produces
- Text
- Open weights
- License: Modified MIT LicenseRuns in about 2463.4 GB of GPU memoryRuntimes: SGLang, vLLM
Cheaper alternatives
Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.
Same model, cheaper host
No host sells it for less than Vercel AI Gateway.
Smaller Moonshot AI models
No cheaper Moonshot AI model keeps its key capabilities.
Other frontier labs
- 12% less
- 52% less
- 79% less
- 79% less
Open-weight picks
- 9% less
- 11% less
- 16% less
- 18% less
Benchmarks
Scores published on the model card.
- Aider Polyglot59.1%
Model line
Newer and older versions, at each lab's own price.
Call it
Using Vercel AI Gateway's API.
from litellm import completion
response = completion(
model="vercel_ai_gateway/kimi-k2",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)Quick answers
How much does Kimi K2 cost?
On Vercel AI Gateway, Kimi K2 costs $0.55 per million input tokens and $2.20 per million output tokens.
Which APIs offer Kimi K2?
Kimi K2 runs on 2 APIs: Vercel AI Gateway and Novita AI.
What is Kimi K2's context window?
131K tokens, with up to 16K tokens in one reply.
Is Kimi K2 being retired?
No retirement date is scheduled for it on any API we track.
Is there a newer version of Kimi K2?
Yes. Kimi K3.1 is the newest model in this line.
What are cheaper alternatives to Kimi K2?
Models that cost less and keep most of what it can do: Gemini 3.5 Flash Lite on Google AI Studio (12% less), GPT-5.4 Nano on OpenAI (52% less) and Motif 3 on ai& (9% less). Savings are on a mix of 3 input tokens to 1 output token.
When was Kimi K2 released?
Moonshot AI released it on 11 Jul 2025.
What is Kimi K2's API model id?
Use kimi-k2 with Vercel AI Gateway. Other hosts use their own ids: kimi-k2-0905 on Novita AI.
Sources
Raw catalog entry for kimi-k2 on Vercel AI Gateway
{
"input_cost_per_token": 5.5e-7,
"litellm_provider": "vercel_ai_gateway",
"max_input_tokens": 131072,
"max_output_tokens": 16384,
"mode": "chat",
"output_cost_per_token": 0.0000022,
"supports_function_calling": true,
"supports_tool_choice": true
}