Skip to main content
Van Data Team
Current modelDeepSeek

DeepSeek-R1 benchmarks, pricing and specs

DeepSeek-R1 is a DeepSeek model released on January 20, 2025 that takes text input and produces text. Its price is not published and accepts up to 128,000 tokens of context. It scores 1398 Elo on Arena Text, #37 of 39 current models (Arena, measured October 8, 2026).

Context window
128K
Arena Text
1398 Elo · #37/39

Data verified October 12, 2026 · Dataset 2026-10-12

Share this

Key facts

Provider
DeepSeek
Model family
DeepSeek-R1
Release date
January 20, 2025
Status
Current
Access
API and open weights
API model id
Not reported
Context window
128,000 tokens
Max output
Not reported
Input
text
Output
text
Task families
text
Declared capabilities
reasoning

DeepSeek-R1 pricing

UnitPrice (USD)TierAppliesSource
1M input tokensNot reportedstandardCurrent
1M output tokensNot reportedstandardCurrent
1M input tokensNot reportedcachedCurrent

Independent benchmark results

Scores come from independent evaluators that ran the model themselves, pinned to a dated snapshot. Rank is among current models in the same snapshot. DeepSeek's own launch figures are listed under strengths with the publisher named, because they are not measured the same way.

BenchmarkScoreRankMeasuredSource
Arena Text1398 Elo#37 of 39October 8, 2026Arena · 2026-10-12
GPQA Diamond71.7%#34 of 37October 12, 2026Epoch AI · 2026-10-12
LiveBench Web of Lies V2100 % correctNot rankedNovember 25, 2024LiveBench · 2024-11-25

Strengths, cautions and best uses

Strengths

  • DeepSeek-R1 records both API and open-weight access for text inputs and a published 128,000-token context window

Cautions

  • Hosted token price and infrastructure cost are not represented; measure deployment and operating cost directly

Best for

  • Reasoning trials requiring API or open-weight deployment options

Alternatives to DeepSeek-R1

The closest current textmodels by price. The cheapest model that scores as well or better is listed first; failing that, the cheapest one within 3% of the benchmark's scale.

Compare DeepSeek-R1 with

DeepSeek-R1 FAQ

What is the context window of DeepSeek-R1?

DeepSeek-R1 accepts up to 128,000 tokens of context. That ranks #40 of 44 current text models by context length.

How does DeepSeek-R1 score on independent benchmarks?

DeepSeek-R1 scores 1398 Elo on Arena Text, #37 of 39 current models (Arena, measured October 8, 2026), 71.7% on GPQA Diamond, #34 of 37 current models (Epoch AI, measured October 12, 2026) and 100 % correct on LiveBench Web of Lies V2 (LiveBench, measured November 25, 2024). These are independent evaluations, not DeepSeek's own launch figures.

Is DeepSeek-R1 open source or API only?

DeepSeek-R1 is available both through DeepSeek's API and as open weights you can run yourself.

What are the best alternatives to DeepSeek-R1?

The closest current alternatives by price and measured performance are Gemini 2.5 Pro (47 points higher on Arena Text, 1445 vs 1398), GPT-4.1 (17 points higher on Arena Text, 1415 vs 1398), GPT-4.1 mini (15 points lower on Arena Text, 1383 vs 1398) and Qwen2.5-VL 72B Instruct (from Alibaba Qwen).

Sources

Cite this page

Van Data Team. "DeepSeek-R1 Benchmarks, Pricing and Specs." AI Model Benchmark Dataset 2026-10-12. https://vandatateam.com/ai-model-benchmarks/deepseek-r1

Reuse the data: JSON · CSV · or explore every model in the AI model benchmark comparison.

Production evaluation

Deciding whether DeepSeek-R1 fits your workload?

Public benchmarks narrow the shortlist. We test the finalists on your own tasks, data and cost limits, then build the routing and review controls to run the winner in production.

  • Evaluation cases and acceptance thresholds tied to real work
  • Model routing, fallbacks and cost controls
Compare all modelsDiscuss your evaluation