Model Detail Human reviewed License review needed

Kimi K3

Moonshot AI Kimi K3 is an open-weight 2.8T-parameter multimodal MoE model with 104B active parameters, native MXFP4 weights, always-on reasoning, tool use, and a 1M-token context window. The official release targets large multi-accelerator deployments through vLLM, SGLang, or TokenSpeed rather than consumer local hardware.

Moonshot AIKIMIkimi-k32026-07-27
Parameters
2800B
104B active
Context window
1M
Long context
Architecture
kda gated mla stable latent moe
moe
Quality score
95
Planner signal

Task Fit

Code AgentSupported

Tool use, repo work, terminal workflows, and coding benchmarks.

CodeSupported

Code generation, debugging, refactoring, and benchmark signal.

ChatSupported

General writing, Q&A, and assistant use.

RAGSupported

Document QA benefits from long context and instruction following.

VisionSupported

Image or visual understanding, not necessarily image generation.

Image GenerationNot a fit

Not marked for image generation in the current library.

Video GenerationNot a fit

Not marked for video generation in the current library.

VoiceNot a fit

Not marked for voice in the current library.

Source Confidence

Overallhigh · 94/100
ParametersReviewed / seeded
Task fitReviewed / seeded
MemoryEstimated from artifact
LicenseNeeds review
BenchmarksAvailable
Hardware fitCalculated
Review flags
memory estimated

Variants and Quant Artifacts

Choose the artifact first; hardware fit follows from RAM, VRAM, format, and runtime.

1 artifacts
QuantFormatQualityMin RAMReco RAMRuntimeAction
MXFP4safetensorsbalanced2TB2TBtransformers, vllm, sglang, tokenspeed Plan with this

Recommended Hardware

No compatible local hardware found in the current hardware pool.

Benchmarks

GPQA Diamond93.5%
Terminal Bench88.3%
MMMU Pro81.6%
Program Bench77.8%
Deep SWE67.5%

Source and Review

Hugging Facemoonshotai/Kimi-K3
OllamaNot mapped
VerificationHuman reviewed
Artifact sourceauto-estimated-from-hf-seed
Default variantKimi K3 Multimodal
Tool callingSupported

Similar Models

Continue planning your local AI setup