ai.canshu.com
中文
VRAM Calculator Advisor & TCO API Pricing Compare Models Leaderboard GPUs

Search models, GPUs, and tools

↑↓ Navigate · Enter Open · ★ Watchlist prioritized ai.canshu.com
Moonshot Kimi
Open-Weight Moonshot Kimi · MoE 2.8T / A104B · Text+Image+Video · MXFP4 / MXFP8 QAT

Kimi K3 (2.8T MoE) Architecture Parameters, Benchmarks & VRAM

World’s first 2.8T-parameter (104B active) open-weight multimodal agentic model with KDA + Gated MLA.

🧬 Kimi K3 (2.8T MoE) Architecture & Tensor Specifications

Verified directly against HuggingFace (moonshotai/Kimi-K3) config.json & Model Card

HF Class: KimiK3ForConditionalGeneration

1. Backbone & Dimensions

kimi_k3
Total / Active Params
2.8T / 104B
Hidden Size (d_model)
7,168
FFN Intermediate Size
33,792
MoE Expert Dim
3,072
MoE Experts
896 Routed · Top-16 + 2 Shared
Activation / Norm
SiTU-GLU · ε=1e-5
Native Tensor Dtype
MXFP4 / MXFP8 QAT

2. Attention & Layer Layout

MLA
Num Hidden Layers
93 Layers (24 Global KV)
Hidden Layer Layout
93 Layers: 69 Kimi Delta Attention (KDA) + 24 Gated MLA + Attention Residuals
Attention Heads (Q / KV)
96 Q / 96 KV
Head Dimension
64
Linear / KV Compression
896 routed + 2 shared experts (16 selected/token) · MoonViT-V2 (401M) vision encoder

3. Context, RoPE & Tokenizer

Tokenizer
Native Context Length
1,048,576 tokens
RoPE Configuration
KDA + Gated MLA RoPE (原生 1,048,576 / 1M Context)
Vocab Size
163,840 (Untied)
Special Token IDs
BOS: 163584 · EOS: 163586 · PAD: 163839

📋 Kimi K3 (2.8T MoE) Official API & System Card Specifications

Verified against official Moonshot Kimi developer documentation & pricing

API Model ID: kimi-k3

1. Identity & Capabilities

API Spec
API Model ID
kimi-k3
Release Date
2026-07-16
Weight License
Open-Weight (API Spec)
Native Reasoning
✅ Supported
Function Calling & JSON Schema
✅ 原生支持
Batch API 异步批处理
✅ Supported (50% off)

2. Context, Output & Modalities

Limits
Max Context Window
1,024,000 tokens (1024K)
Max Output Tokens
1,049,000 tokens
Input Modalities
Text + Image + Video
Output Modalities
Text
Direct China Access
✅ Direct CN + RMB
Deployment Mode
Open-Weight + Cloud API

3. Official Pricing & Cache

CNY / 1M
Input / 1M Tokens
¥ 20
Cached Input / 1M
¥ 2 -90%
Output / 1M Tokens
¥ 100
Tiered Pricing Notes
Moonshot AI: Moonshot AI official CN platform with auto prefix cache

🏆 Official Benchmark Results

Sourced from official System Card / Model Card & technical reports

Arena Elo ~1515 AA Index 54
🧠 Reasoning
GPQA Diamond 93.5 %

PhD-level Google-proof science reasoning questions

🧠 Reasoning
HLE-Full (w/ Tools) 56 %

Full multimodal Humanity's Last Exam with tools

📚 Knowledge
AA-LCR 74.7 %

Agent Arena Long Context Reasoning benchmark.

💻 Coding
DeepSWE v1.1 67.5 %
🤖 Agentic
CyberGym 80 %

CyberGym is a benchmark for evaluating AI agents on cybersecurity tasks, testing their ability to identify vulnerabilities, perform security analysis, and complete security-related challenges in a controlled environment.

👁️ Multimodal
BabyVision (w/ Tools) 85.7 %
👁️ Multimodal
CharXiv RQ 84.8 %

Scientific chart synthesis and quantitative reasoning

🎛️ Official Sampling & Reasoning Parameters

Recommended generation settings & API parameter constraints

Max Effort Reasoning & Agentic
temperature=1 top_p=0.95 max_tokens=65,536

💡 Use top_p=0.95 for single-step reasoning and top_p=1.0 for multi-step agentic harnesses.

📊 Kimi K3 (2.8T MoE) Full Benchmark Scorecard

29 public benchmarks · ranks among 154 models · synced 2026-11-20

⚔️ Compare head-to-head

💻 Coding 10 items

Terminal-Bench 2.1 #6/37 88.3%

Terminal-Bench 2.1 is an updated release of the Terminal-Bench benchmark that tests AI agents' ability to operate a computer via the terminal.

MCP Atlas #5/39 84.2%

MCP Atlas is a benchmark for evaluating AI models on scaled tool use capabilities, measuring how well models can coordinate and utilize multiple tools across complex multi-step tasks.

FrontierSWE #2/16 81.2%

FrontierSWE measures whether an agent can complete open-ended technical projects at the scale of hours to tens of hours, spanning systems optimization, large-scale code construction, and applied ML research. Performance is reported as a dominance score, where higher is better.

Program Bench #5/18 77.8%

Program Bench evaluates code-generation agents by asking them to recreate a program's behavior from only a compiled binary and documentation.

Kimi Code Bench v2 #2/3 72.9%

Kimi Code Bench v2 is Moonshot AI's in-house benchmark for evaluating coding agents on realistic software engineering tasks across 10+ mainstream programming languages and a production tech stack spanning backend services, infrastructure, performance engineering, systems…

DeepSWE 1.1 #19/49 69%

DeepSWE 1.1 evaluates software engineering agents using the mini-swe-agent harness.

DeepSWE #4/12 67.5%

DeepSWE is a software engineering agent benchmark evaluated with the mini-swe-agent harness, where each task is solved in an isolated container with no internet access. It measures an agent's ability to autonomously resolve real-world coding issues end to end.

MLS-Bench Lite #1/2 48.3%

MLS-Bench Lite is the official 30-task subset of MLS-Bench for evaluating whether AI systems can invent generalizable and scalable machine learning methods across LLM pretraining and post-training, robotics, world models, computer vision, reinforcement learning, optimization, ML…

SWE-Marathon #5/9 42%

SWE-Marathon is an ultra-long-horizon software engineering benchmark covering tasks such as building compilers, optimizing kernels, and developing production-grade services. It measures whether agents can sustain quality across extremely long engineering trajectories.

PostTrainBench #5/10 36.6%

PostTrainBench evaluates a model's ability to autonomously post-train base models. Given pretrain-only base models, the agent must complete the full pipeline of data synthesis, training, evaluation, and iteration within a time budget, scored across downstream benchmarks such as…

🤖 Agents & Tools 9 items

DeepSearchQA #1/8 95%

DeepSearchQA is a benchmark for evaluating deep search and question-answering capabilities, testing models' ability to perform multi-hop reasoning and information retrieval across complex knowledge domains.

BrowseComp #4/54 91.2%

BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information.

Toolathlon #6/32 73.2%

Tool Decathlon is a comprehensive benchmark for evaluating AI agents' ability to use multiple tools across diverse task categories. It measures proficiency in tool selection, sequencing, and execution across ten different tool-use scenarios.

OfficeQA Pro #9/15 63.3%

OfficeQA Pro evaluates AI models on professional knowledge-work questions and tasks drawn from real office workflows, including document analysis, spreadsheet reasoning, and information synthesis across business domains.

Job Bench #8/10 52.9%

Job Bench evaluates AI agents on realistic professional tasks that require multi-step planning, research, and production of work artifacts.

AA-Briefcase #2/5 1548

AA-Briefcase is an Artificial Analysis evaluation of AI systems on professional knowledge-work tasks, reported as an Elo score.

APEX-Agents #7/15 37.6%

APEX-Agents is a benchmark evaluating AI agents on long horizon professional tasks that require sustained reasoning, planning, and execution across complex multi-step workflows.

SpreadsheetBench 2 #3/3 34.8%

SpreadsheetBench 2 evaluates office automation agents on spreadsheet analysis, reasoning, and manipulation tasks.

AutomationBench #17/26 30.8%

AutomationBench is a tool-use benchmark that evaluates AI agents on automating real-world workflows, testing their ability to orchestrate tools and complete multi-step automation tasks.

Source: ZeroEval / llm-stats.com (✓ = third-party verified, others vendor-reported).

⚡ Kimi K3 (2.8T MoE) Interactive VRAM & Token Speed Calculator

Concurrency (Batch):
MoE 2800B / A104B·93L (d=7168)·96Q/96KV·Ctx 1M·Text+Image+Video
Runnable Models →
GPUs
96 GB VRAM·4000 GB/s·Safe Usable 94.1 GB
Out of VRAM (OOM)1733.2 / 94.1 GBOOM
3. Weight Quantization
4. Per-Request Context16K (16,384 tok)
Verified: 2026-11-20·Roofline bandwidth/compute estimate (Batch=1)
Out of VRAM (OOM)
1733.2/ 94.1 GB usable

Physical 96 GB · Load 1842.3%

Theoretical Decode Speed
—tokens/s

Exceeds usable VRAM by 1639.12 GB

Max Context at Batch=1
0K/ 1M

MLA compresses multi-batch KV footprint

Weights: 1697.5 GBKV Cache (1×16K): 0.45 GBOverhead: 35.25 GB
⛔ Cannot Load
Reference Launch CLI:
OLLAMA_CONTEXT_LENGTH=16384 ollama run kimi-k3:latest

Insufficient Local VRAM? Ways to Run Kimi K3 (2.8T MoE)

Rent an hourly high-VRAM cloud GPU pod, or check the 12-month Buy vs Cloud vs API break-even curve.

📊 Kimi K3 (2.8T MoE) Quantization VRAM Breakdown (16K Context)

Quantization Bits/Weight Weights KV Cache (16K) Total VRAM Quality Note
FP16 / BF16 16 bpw 5600 GB 0.45 GB 5713.75 GB Lossless Full Precision
FP8 / Q8_0 8.5 bpw 2975 GB 0.45 GB 3036.25 GB 99.9% Near-Lossless
GGUF Q6_K 6.56 bpw 2296 GB 0.45 GB 2343.67 GB Ultra High Quality
GGUF Q5_K_M 5.69 bpw 1991.5 GB 0.45 GB 2033.08 GB High Balanced Quality
Q4_K_M / AWQ 4.85 bpw 1697.5 GB 0.45 GB 1733.2 GB Sweet Spot (Recommended)
GGUF IQ4_XS 4.25 bpw 1487.5 GB 0.45 GB 1519 GB Compact 4-bit
GGUF Q3_K_M 3.91 bpw 1368.5 GB 0.45 GB 1397.62 GB Moderate Perplexity Loss
GGUF Q2_K 3.15 bpw 1102.5 GB 0.45 GB 1126.3 GB Extreme Compression

🖥️ GPU & Apple Silicon Compatibility for Kimi K3 (2.8T MoE) (Q4_K_M)

Frequently Asked Questions

What are the exact architecture parameters of Kimi K3 (2.8T MoE)?

Kimi K3 (2.8T MoE) uses KimiK3ForConditionalGeneration with 93 layers (93 Layers: 69 Kimi Delta Attention (KDA) + 24 Gated MLA + Attention Residuals), hidden_size=7168, 96 Q heads / 96 KV heads (head_dim=64), and vocab_size=163,840.

How much VRAM does Kimi K3 (2.8T MoE) require?

At 4-bit quantization (Q4_K_M, 16K context), Kimi K3 (2.8T MoE) requires 1733.2 GB VRAM. Full FP16 requires 5713.75 GB.