How much does the Kimi K2 Thinking API cost?

Kimi K2 Thinking costs $0.6 per million input tokens and $2.5 per million output tokens, with cached input at $0.15 per million. A workload of 10M input and 1.5M output tokens per month costs about $9.75.

What is the context window of Kimi K2 Thinking?

Kimi K2 Thinking supports a context window of 262,144 tokens (262K).

Is Kimi K2 Thinking open source?

Yes — Kimi K2 Thinking is an open-weights model released under the Modified MIT license, so it can be downloaded and self-hosted.

What are the best alternatives to Kimi K2 Thinking?

The closest alternatives by overall capability are DeepSeek V3.2, OpenAI o3, Claude Sonnet 4.5, GPT-5. See the comparison pages for detailed head-to-head breakdowns.

Kimi K2 Thinking

Name: Kimi K2 Thinking
Price: 0.6 USD
Author: Moonshot AI

Moonshot AI·Nov 2025reasoningopen weights · Modified MIT

67.7

modhub Index

Kimi K2 Thinking stunned the field in November 2025: an open-weights trillion-parameter MoE (32B active) that matched or beat closed flagships on agentic benchmarks like Humanity's Last Exam with tools, and could sustain hundreds of sequential tool calls without drifting. Trained for a reported $4.6M, it became the strongest argument that open models had caught the closed frontier. It remains a top choice for self-hosted deep-research and agent pipelines.

Benchmark results

AIME 2025Math · #4 of 16~94.5%
MMLU-ProKnowledge · #6 of 11~84.6%
GPQA DiamondReasoning · #10 of 2684.5%
SWE-bench VerifiedCoding · #20 of 3371.3%
Terminal-BenchAgentic · #6 of 10~47.1%
HLEReasoning · #6 of 1423.9%

Where it shines

Elite agentic search and tool-use endurance
Open weights at trillion-parameter scale
INT4 quantization-aware training for practical serving

Alternatives to Kimi K2 Thinking

DeepSeek V3.2

The sparse-attention release that halved DeepSeek's already-lowest-in-class prices while keeping GPT-class quality.

OpenAI o3

OpenAI's dedicated 2025 reasoning model that pioneered thinking-with-images and agentic tool use within chain-of-thought.

Claude Sonnet 4.5

The model that made 30-hour autonomous coding sessions real — 77.2% SWE-bench Verified and a 1M-token context beta at $3/$15.

GPT-5

OpenAI's August 2025 unified flagship that merged the GPT and o-series reasoning lines into one model with selectable effort.

Frequently asked questions

How much does the Kimi K2 Thinking API cost?: Kimi K2 Thinking costs $0.6 per million input tokens and $2.5 per million output tokens, with cached input at $0.15 per million. A workload of 10M input and 1.5M output tokens per month costs about $9.75.
What is the context window of Kimi K2 Thinking?: Kimi K2 Thinking supports a context window of 262,144 tokens (262K).
Is Kimi K2 Thinking open source?: Yes — Kimi K2 Thinking is an open-weights model released under the Modified MIT license, so it can be downloaded and self-hosted.
What are the best alternatives to Kimi K2 Thinking?: The closest alternatives by overall capability are DeepSeek V3.2, OpenAI o3, Claude Sonnet 4.5, GPT-5. See the comparison pages for detailed head-to-head breakdowns.