Voltar para Explorar

DeepSeek R1 Distill Qwen 14B

DeepSeek·Lançado em 2025-01-20

Open Source14BMITMoE

Comparar Ver repositório do modelo

Modelos Relacionados

DeepSeek V4 Pro (Reasoning, High Effort)2026-04-24 DeepSeek V4 Flash (Reasoning, High Effort)2026-04-24 DeepSeek V4 Pro (Non-reasoning)2026-04-24 DeepSeek V4 Flash (Non-reasoning)2026-04-24 DeepSeek: DeepSeek V4 Flash (free)2026-04-24 DeepSeek V4 Pro2026-04-22 DeepSeek V4 Flash2026-04-22 DeepSeek V3.2 (Reasoning)2025-12-01

Benchmarks

MMLU-Pro

74.0%

GPQA Diamond

48.4%

HLE

4.4%

LiveCodeBench

37.6%

SciCode

23.9%

TerminalBench HardNão avaliado

MATH-500

94.9%

AIME

66.7%

AIME 2025

55.7%

IFBench

22.1%

Long Context Recall

7.0%

Tau2Não avaliado

Média do MercadoMelhor Score

Open Source

Ver repositório do modelo

mit14BGGUF / GPTQ / AWQ

Downloads

710.4K

Curtidas

610

VRAM (FP16)

16-24 GB

GPU

RTX 4090 / M2 Max

Comparação Rápida

Modelos Similares

Falcon-H1R-7B

TII

DeepSeek: R1 Distill Llama 70B

DeepSeek

Q: 9.9$0.80/1M128K ctx

Claude 3.5 Sonnet (Oct '24)

Anthropic

Ling-flash-2.0

InclusionAI

Qwen3 VL 30B A3B Instruct

Alibaba

Q: 10.0$0.23/1M262K ctx

Meta: Llama 4 Scout

Meta

Q: 10.0$0.15/1M10.0M ctx

Comparar todos os 7 modelos