DeepSeek Reasoner

DeepSeek · deepseek/deepseek-reasoner

ProviderDeepSeek
Familydeepseek-thinking
Typellm-reasoning
Statusactive
Released2025-12-01
Updated2026-02-28

Reported benchmark scores

These scores carry one collection date for the whole card (2026-04) and a source list rather than a source per score, so they cannot be attributed individually. They are shown as reported and are not verified evidence. Benchmark pages carry the reviewed, dated evidence where it exists.
BenchmarkCatalogue standingScore
Aider Polyglot Benchmarkunassessed65.8
AIME 2025unassessed79.8
AlpacaEvalunassessed45.2
Arena Elo — Codingunassessed1370.0
Arena Elo — Mathunassessed1400.0
Arena Elo — Overall (Text)unassessed1360.0
BIG-Bench Hardunassessed68.7
GPQA Diamondunassessed71.5
GSM8Kunassessed89.3
HumanEvalunassessed89.2
IFEvalunassessed86.5
LiveCodeBenchunassessed52.1
MATH-500unassessed97.3
MMLU: Astronomyunassessed76.5
MMLU: Biology (subcategory)unassessed84.2
MMLU: Business Ethicsunassessed77.8
MMLU: Chemistry (subcategory)unassessed78.5
MMLU: Clinical Knowledgeunassessed83.5
MMLU: Computer Science (subcategory)unassessed85.5
MMLU: Jurisprudenceunassessed74.5
MMLU: Physics (unresolved key)unassessed82.8
MMLU-Prounassessed79.8
MMLU: Professional Accountingunassessed68.2
MMLU: Professional Lawunassessed72.5
MT-Benchunassessed9.0
MultiPL-Eunassessed84.6
MultiPL-E: C++unassessed83.5
MultiPL-E: C#unassessed80.5
MultiPL-E: Gounassessed80.5
MultiPL-E: Javaunassessed87.2
MultiPL-E: JavaScriptunassessed86.8
MultiPL-E: Juliaunassessed62.1
MultiPL-E: Kotlin (unconfirmed)unassessed72.5
MultiPL-E: Luaunassessed55.8
MultiPL-E: Perlunassessed50.2
MultiPL-E: PHPunassessed77.8
MultiPL-E: Pythonunassessed92.8
MultiPL-E: Runassessed58.5
MultiPL-E: Rubyunassessed65.5
MultiPL-E: Rustunassessed75.1
MultiPL-E: Scalaunassessed60.2
MultiPL-E: Swiftunassessed65.2
MultiPL-E: TypeScriptunassessed86.1
MuSRunassessed33.5
SWE-bench Agentunassessed42.5
SWE-bench Verifiedunassessed49.2
τ-benchunassessed48.8
WildBenchunassessed72.5

Data

This card as JSON · Edit on GitHub