Kimi K2 Thinking

Language Moonshot AI Partly verified Released Nov 2025
  • In Input modalities: Text Out Output modalities: Text
  • Reasoning
  • Tool use
  • Structured output
  • Open weights

Prices

3 offerings · prices as listed
Alibaba Cloud Model Studio (opens in a new tab) Official Available on this channel since Nov 10, 2025 China · CNY Context: 262.14KReasoning: YesTool use: YesStructured output: YesMax input: 229.38KMax output: 16.38KInput modalities: TextOutput modalities: TextBuilt-in tools: Web search kimi-k2-thinking
Conditions/ 1M tokensInputOutput
Standard no longer listed since Oct 11, 2026 ¥4.00¥16.00

Prices from 阿里云百炼:模型价格 (opens in a new tab) (checked Oct 8, 2026)

Moonshot AI Platform (opens in a new tab) Official Global · USD kimi-k2-thinking
Conditions/ 1M tokensInputCache readOutput
Standard past price, from the vendor’s page archived Mar 30, 2026$0.60$0.15$2.50

Prices from Moonshot AI platform pricing (opens in a new tab) (checked Oct 23, 2026)

Moonshot AI Platform (opens in a new tab) Official China · CNY kimi-k2-thinking
Conditions/ 1M tokensInputCache readOutput
Standard past price, from the vendor’s page archived Feb 17, 2026¥4.00¥1.00¥16.00

Prices from Kimi API 开放平台:模型推理价格 (opens in a new tab) (checked Oct 23, 2026)

Standard: the price a request with no conditions pays (the headline price elsewhere on the site).

Details

Max input
229.38K
Built-in tools
Web search
Knowledge cutoff
Aug 2024
Family
kimi-k-thinking
Model ID

Not stated on Moonshot AI’s pages: reasoning efforts, reasoning can be off, max thinking budget, aliases.

Run locally

Weights: moonshotai/Kimi-K2-Thinking
  • Parameters 1.03T
  • Active 32B · 384 experts, 8 per token
  • Native precision INT4 (AWQ / GPTQ)
  • Layers 61
  • KV cache per token 69 KB
  • Config positions 256K (config.json max_position_embeddings)
  • API context 250K (maker’s API)
My hardware GeForce RTX 4090 24 GB · 1008 GB/s · 64 GB DDR5 dual channel · 89.6 GB/s
On your hardware · GeForce RTX 4090
Doesn’t fit INT4 (AWQ / GPTQ) (native) (automatic) 8K context
WeightsKV cacheOverheadUsable
Needs
≈ 581.6 GB (weights 553.4 (maker’s file) + KV cache 0.5 + overhead 27.7)
Usable
21.6 GB (90% of its memory)
Try
it’s too large for this hardware; see the smallest that runs it below
PrecisionWeightsNeeds at 8KOn your hardwareLongest context
INT4 (AWQ / GPTQ) (native)maker’s file553.4 GB≈ 581.6Doesn’t fit—
Q3_K_Mquality drops467.2 GB≈ 491.1Doesn’t fit—
IQ3_XXSquality drops365.6 GB≈ 384.5Doesn’t fit—
Q2_Kquality drops358.5 GB≈ 376.9Doesn’t fit—
IQ2_XXSquality drops246.1 GB≈ 259.0Doesn’t fit—

Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.

Compared with other hardware

Each at the highest precision it runs. Click one for details.

Smallest that runs it (INT4 (AWQ / GPTQ) (native), 8K): no single listed machine; about 9 × 80 GB GPUs.

Run commands

vLLM
vllm serve moonshotai/Kimi-K2-Thinking --max-model-len 8192
Memory to fine-tune
Full≈ 15298.6 GBDoesn’t fit your hardware
LoRA≈ 1916.6 GBDoesn’t fit your hardware
QLoRA≈ 542.4 GBDoesn’t fit your hardware

Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.

Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-21. How it’s estimated

Cited scores

From third-party leaderboards, not measured by us
  • Epoch Capabilities Index 145.5

    #78 of 123 listed models · CI 143.7–147.3

  • EQ-Bench Creative Writing v3 1,631

    #38 of 95 listed models

  • EQ-Bench Longform Writing 69

    #41 of 93 listed models

  • Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Nov 14, 2026
  • EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · License · as of Oct 28, 2026
  • changes we made
  1. Now priced on Moonshot AI Platform, China Cache read ¥1.00 · Input ¥4.00 · Output ¥16.00
  2. Now priced on Moonshot AI Platform Cache read $0.15 · Input $0.60 · Output $2.50
  3. Now priced on Moonshot AI Platform, China Cache read ¥1.00 · Input ¥4.00 · Output ¥16.00
  4. Now priced on Moonshot AI Platform Cache read $0.15 · Input $0.60 · Output $2.50
  5. Now priced on Alibaba Cloud Model Studio, China Input ¥4.00 · Output ¥16.00
  6. Context window 256K → 262.14K

Sources

Moonshot AI’s official policies: data use, privacy, terms →

Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data

Moonshot AI · China · moonshot.ai

Snapshots, the same family and other models from this provider