HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks
Home/Orgs/moonshotai/moonshotai/Kimi-K2-Thinking

moonshotai/Kimi-K2-Thinking

family kimi · 5 variants · unknown

provider
moonshotai
released
—
huggingface
https://huggingface.co/moonshotai/Kimi-K2-Thinking
best scores
Humanity's Last Exam: 23.9AIME 2025: 94.5GPQA: 84.5MMLU-Pro: 84.6MMLU: 94.4BrowseComp: 62.3SWE-bench Verified: 71.3SciCode: 44.8LiveCodeBench: 83.1Terminal-Bench: 47.1MT-Bench: 2.28MATH: 3.34

Best scores per benchmark

benchmarkcategorybest score
AIME 2025math94.5
MMLUknowledge94.4
MMLU-Proknowledge84.6
GPQAreasoning84.5
LiveCodeBenchcoding83.1
SWE-bench Verifiedswe71.3
BrowseCompagentic62.3
Terminal-Benchagentic-coding47.1
SciCodecoding44.8
Humanity's Last Examknowledge23.9
MATHmath3.34
MT-Benchchat2.28

Variants (5)

variantkindquantbitsvRAMdlorgyear
Kimi-K2-Thinkingoriginalbf1616.073,772moonshotai2026
Kimi-K2-Thinking-NVFP4nvfp4nvfp44.013,139nvidia2026
Kimi-K2-Thinking-GGUFggufgguf16.01,398unsloth2026
Kimi-K2-Thinkingoriginalbf1616.071unsloth2026
Kimi-K2-Thinking-Eagle3originalbf1616.066nvidia2026
Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.