Model creator

DeepSeek

Every DeepSeek model in the BenchSift catalogue, with its scores, ranks, prices and capabilities as the sources publish them.

View in the catalogue
Models
36
48 configurations, reasoning levels included
Open weights
36
36 of 36 models
Latest release
Oct 17, 2026
DeepSeek V4.1 Flash 1017 (High) · First dated model: Nov 29, 2023
Median price
$0.521
$0.142 to $2.50 per million tokens, across 18 priced models

Best scores and ranks

The creator's best model on each measure, and its rank among every measured model in the catalogue.
34.8Rank 81 of 782 measured models
69.1Rank 58 of 301 measured models
96.7Rank 4 of 270 measured models
47.5Rank 24 of 191 measured models
203 t/sRank 29 of 449 measured models
$0.117Rank 126 of 740 measured models
#1

Capabilities

Number of models that declare each capability. Tool calling and structured outputs come from OpenRouter.
Reasoning25 of 36
Open weights36 of 36
Tool calling11 of 36
Structured outputs11 of 36

Input

  • Text36
  • Image5

Output

  • Text35
  • Image1

Overview

Model types

  • Text35
  • Image1

Releases by year

  • 202611
  • 202517
  • 20246
  • 20231

Hugging Face

Official repositories confirmed by Hugging Face metadata.

1 official repository

Licenses

  • mit1

Models

36

Reasoning levels are folded into one entry. The rank is the whole catalogue's: sorting does not change it.

  • DeepSeek V4 Flash Vision (Max)

    Text · Reasoning · Open weights · Aug 21, 2026 · $0.66 · 141 t/s

    34.8#81 of 782
  • DeepSeek V4 Flash 0731 (Max)

    Text · Reasoning · Open weights · Jul 31, 2026 · $0.175 · 129 t/s

    34.3#89 of 782
  • DeepSeek V4.1 Flash 1017

    Text · Reasoning · Open weights · 3 configurations · Oct 17, 2026 · $0.525 · 152 t/s

    32.6#103 of 782
  • DeepSeek V4 Pro 0424

    Text · Reasoning · Open weights · 3 configurations · Apr 24, 2026 · $1.77 · 71.9 t/s

    30.4#123 of 782
  • DeepSeek V4.1 Flash

    Text · Reasoning · Open weights · 2 configurations · Sep 10, 2026 · $0.517 · 203 t/s

    28.6#141 of 782
  • DeepSeek V4 Flash 0420

    Text · Reasoning · Open weights · 3 configurations · Apr 24, 2026 · $0.142 · 58.3 t/s

    24.2#198 of 782
  • DeepSeek V4 Pro 0813

    Text · Reasoning · Open weights · 2 configurations · Aug 13, 2026 · $1.98 · 75 t/s

    23.1#214 of 782
  • DeepSeek V3.2

    Text · Reasoning · Open weights · 2 configurations · Dec 1, 2025 · $0.315 · 46.3 t/s

    21.5#239 of 782
  • DeepSeek V3.1 Terminus

    Text · Reasoning · Open weights · 2 configurations · Sep 22, 2025 · $0.453 · 71.5 t/s

    14.8#332 of 782
  • DeepSeek V3.2 Speciale

    Text · Reasoning · Open weights · Dec 1, 2025

    14.5#336 of 782
  • DeepSeek V3.2 Exp

    Text · Reasoning · Open weights · 2 configurations · Sep 29, 2025 · $0.315

    13.9#346 of 782
  • DeepSeek V3.1

    Text · Reasoning · Open weights · 2 configurations · Aug 21, 2025 · $0.84 · 70.4 t/s

    13.7#349 of 782
  • DeepSeek R1 0528 (May '25)

    Text · Reasoning · Open weights · May 28, 2025 · $2.08 · 33.1 t/s

    13.1#363 of 782
  • R1 0528

    Text · Reasoning · Open weights · May 28, 2025 · $0.913

    13.1#363 of 782
  • DeepSeek R1 (Jan '25)

    Text · Reasoning · Open weights · Jan 20, 2025 · $2.50 · 129 t/s

    11.4#410 of 782
  • DeepSeek V3 0324

    Text · Open weights · Mar 25, 2025 · $0.405 · 125 t/s

    9.7#456 of 782
  • DeepSeek V3 (Dec '24)

    Text · Open weights · Dec 26, 2024 · $0.463 · 75.5 t/s

    8.5#509 of 782
  • DeepSeek R1 Distill Qwen 32B

    Text · Reasoning · Open weights · Jan 20, 2025

    8.4#511 of 782
  • DeepSeek R1 0528 Qwen3 8B

    Text · Reasoning · Open weights · May 29, 2025

    8.1#525 of 782
  • DeepSeek R1 Distill Llama 70B

    Text · Reasoning · Open weights · Jan 20, 2025 · $0.80

    7.9#527 of 782
  • DeepSeek R1 Distill Qwen 14B

    Text · Reasoning · Open weights · Jan 20, 2025

    7.8#533 of 782
  • DeepSeek-V2.5 (Dec '24)

    Text · Open weights · Dec 10, 2024

    6.6#622 of 782
  • DeepSeek-V2.5

    Text · Open weights · Sep 6, 2024

    6.6#622 of 782
  • DeepSeek R1 Distill Llama 8B

    Text · Reasoning · Open weights · Jan 20, 2025

    6.5#631 of 782
  • DeepSeek-Coder-V2

    Text · Open weights · Jun 17, 2024

    6.0#660 of 782
  • DeepSeek R1 Distill Qwen 1.5B

    Text · Reasoning · Open weights · Jan 20, 2025

    5.5#698 of 782
  • DeepSeek-V2-Chat

    Text · Open weights · May 6, 2024

    5.5#698 of 782
  • DeepSeek Coder V2 Lite Instruct

    Text · Open weights · Jun 17, 2024

    5.3#718 of 782
  • DeepSeek LLM 67B Chat (V1)

    Text · Open weights · Nov 29, 2023

    5.3#718 of 782
  • DeepSeek V4 Flash Vision Exp (batch)

    Text · Reasoning · Open weights · Aug 21, 2026 · $0.165

    —
  • DeepSeek V4 Flash 0423

    Text · Reasoning · Open weights · Apr 23, 2026

    —
  • DeepSeek V4 Pro 0423

    Text · Reasoning · Open weights · Apr 23, 2026

    —
  • DeepSeek OCR 2

    Text · Open weights · Jan 27, 2026

    —
  • DeepSeek Chat

    Text · Open weights · Dec 1, 2025

    —
  • DeepSeek Reasoner

    Text · Reasoning · Open weights · Dec 1, 2025

    —
  • Janus Pro

    Image · Open weights

    —
Where these figures come from

Each catalogue model is tied to its creator by the identity Artificial Analysis publishes, or OpenRouter's and Models.dev's for models it does not track. Counts fold the reasoning levels of one model together. Ranks compare each measure with every measured model in the catalogue, configurations included, as model pages do. The price is the published blended price (3 input tokens to 1 output token).

Figures computed by BenchSift from the catalogue: Artificial Analysis, OpenRouter, Models.dev and Hugging Face. A value no model carries is not shown.

Data retrieved on Nov 12, 2026.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy