Which is better, DeepSeek-V4.1-Flash or Kimi K3?

DeepSeek-V4.1-Flash vs Kimi K3

28 SHARED BENCHMARKS

Across 28 shared benchmarks, DeepSeek-V4.1-Flash scores higher on 12 and Kimi K3 on 16. The widest gap is Program Bench, where Kimi K3 scores 77.8 against 20.3. DeepSeek-V4.1-Flash is the cheaper of the two on tracked API pricing ($0.30 against $2.98 per million input tokens).

DEEPSEEKVSMOONSHOT28 SHARED12–16 HEAD-TO-HEAD

BenchmarkDeepSeek-V4.1-FlashMARGINKimi K3

AA Intelligence4060

AA-LCR8488.7

AA-Omniscience-5.319.7

baby_vision89.685.7

critpt14.323.4

cybergym88.180

DeepSWE 1.174.269

gdpval56.659.4

GPQA Diamond90.993.5

HLE39.256

HLE (with tools)63.959.8

livebench_agentic_coding77.362.2

livebench_coding8081.5

livebench_data_analysis79.378.7

livebench_instruction_following7071.4

livebench_language81.285.5

livebench_math93.384.4

livebench_reasoning86.790.7

MMMU-Pro7783.4

nl2repo6458

OmniScience Accuracy46.447.6

OmniScience Non-Hallucination3.549.1

Program Bench20.377.8

scicode51.959.5

Terminal-Bench 2.190.688.3

Terminal-bench 3.03017.4

terminalbenchV4026.812.6

ZeroBench4941

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (direct, moonshot-official), otherwise the lowest tracked offer.

Compare Kimi K3 withALL PAIRINGS →

MoonshotKimi K2.6 Kimi K2.5 Kimi K2 Instruct

Across the fieldGPT-5.5 Claude Opus 4.8 GPT-4o GPT-5.6 Sol Gemini 3.1 Pro GPT-5.4 GPT-5.2 Claude Opus 4.6