Pangripta model

DeepSeek

deepseek/

Jelajahi model saka DeepSeek. Bandhingake panyedhiya sing kasedhiya, tag model lan rega inferensi ing siji panggonan.

DeepSeek ngembangaké modèl umum, pamikiran lan coding. Rilis modèl DeepSeek V3.2 lan V4 nerbitaké bobot modèl kanthi lisensi MIT.

11 modèl Situs web resmi Saring ing Model

Model

Rega saben 1M token

deepseek/deepseek-coder-v2

Model basa kode Mixture-of-Experts open-source sing entuk kinerja sing padha karo GPT4-Turbo ing tugas-tugas spesifik kode.

64 tag model164K konteksTukar token

Rincian model

deepseek/deepseek-coder

DeepSeek Coder minangka model coding sing mampu dilatih ing rong triliun token kode lan basa alami.

102 tag model16K konteksTukar token

Rincian model

deepseek/deepseek-llm

Model basa canggih sing digawe kanthi 2 triliun token dwibasa.

64 tag modelKonteks 4KTukar token

Rincian model

deepseek/deepseek-v2

Model basa Mixture-of-Experts sing kuat, ekonomis, lan efisien.

34 tag model164K konteksTukar token

Rincian model

deepseek/deepseek-v2.5

Versi upgrade saka DeepSeek-V2 sing nggabungake kemampuan umum lan coding saka DeepSeek-V2-Chat lan DeepSeek-Coder-V2-Instruct.

7 tag model164K konteksTukar token

Rincian model

deepseek/deepseek-r1

DeepSeek-R1 minangka keluarga model reasoning terbuka kanthi performa sing nyedhaki model unggulan, kaya O3 lan Gemini 2.5 Pro.

35 tag model131K konteksTukar token

Rincian model

deepseek/deepseek-v3

Model basa campuran ahli (MoE) sing kuat kanthi total 671B parameter lan 37B diaktifake kanggo saben token.

5 tag model164K konteksTukar token

Rincian model

deepseek/deepseek-v3.2

DeepSeek-V3.2 minangka model basa gedhe saka DeepSeek

164K konteks

Rincian model

DeepSeek-V4-Pro iku model saka DeepSeek sing nampa teks lan ngasilake teks. Entri katalog iki ngenali deepseek/deepseek-v4-pro. Jendhela konteks sing wis direview yaiku 1,048,576 token; watesan deployment bisa luwih cilik.

* Kredit sing kasedhiyaKasedhiya saka ShareAI1M konteksSaka €1.568779/M inputSaka €3.137558/M output

Rincian model

deepseek/deepseek-v4-flash

DeepSeek-V4-Flash iku model saka DeepSeek sing nampa teks lan ngasilake teks. Entri katalog iki ngenali deepseek/deepseek-v4-flash. Jendhela konteks sing wis direview yaiku 1,048,576 token; watesan deployment bisa luwih cilik.

* Kredit sing kasedhiyaKasedhiya saka ShareAI1M konteksSaka €0.171303/M inputSaka €0.459815/M output

Rincian model

deepseek/deepseek-v3.1

DeepSeek-V3.1-Terminus iku modèl hibrida sing ndhukung mode mikir lan mode ora mikir.

8 tag model164K konteksTukar token

Rincian model

Builder Program

Pakaryan Kustom

Aplikasi

Sumber Terbuka

Program Provider

Kepiye sampeyan melu

Program Creator

Program Kemitraan

DeepSeek

Model

deepseek/deepseek-coder-v2

deepseek/deepseek-coder

deepseek/deepseek-llm

deepseek/deepseek-v2

deepseek/deepseek-v2.5

deepseek/deepseek-r1

deepseek/deepseek-v3

deepseek/deepseek-v3.2

deepseek/deepseek-v4-pro

deepseek/deepseek-v4-flash

deepseek/deepseek-v3.1