Inference
EN
DeepSeek

DeepSeek V3.1 (Reasoning)

DeepSeek V3.1 (Reasoning) is a reasoning model from DeepSeek in the DeepSeek V3.1 family. 11 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2025-08
RankedEnough results for an index.
IndexAll categories, out of 100.43.0±6.4Median 51.0
CoverageIndex weight that has results.85%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.128KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

684.53B parameters
16-bit
1438.5 GB
8-bit
764.7 GB
4-bit
436.7 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy needs more than 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.1 counted27.6—
CodingCode writing and repair.1 counted51.4—
ReasoningLogic problems and puzzles.2 counted43.5148 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.4 counted47.0157 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted27.8—
MathMath problems.2 counted46.599 / 168

Results

11 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
American Invitational Mathematics Examination 2024Math93.1%3 / 27
LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for CodeCoding74.8%2 / 12
Massive Multitask Language Understanding ProfessionalKnowledge84.8%23 / 75
American Invitational Mathematics Examination 2025Math88.4%20 / 53
Graduate-Level Google-Proof Q&AKnowledge80.1%66 / 141
Critical Physics TasksReasoning2.0%105 / 214
Artificial Analysis Omniscience AccuracyKnowledge29.0%82 / 164
Artificial Analysis GPQA DiamondKnowledge77.9%106 / 199
Artificial Analysis Humanity's Last ExamKnowledge14.3%128 / 222
Artificial Analysis Long Context ReasoningReasoning56.7%144 / 217
Artificial Analysis IFBenchInstruction41.5%98 / 146
τ²-Bench Tool-Agent-User EvaluationAgentic37.4%105 / 149
Artificial Analysis MMLU-ProKnowledge85.1%13 / 15
Humanity's Last ExamKnowledge15.9%70 / 77
Artificial Analysis AIME 2025Math89.7%—

Sources

BenchLMCC BY-NC 4.0

History

5 releases
Index43.8 → 43.0
23 Oct 202612 Nov 2026

Questions

How good is DeepSeek V3.1 (Reasoning)?
Its index is 43.0 out of 100. That is place 193 of 283 ranked models.
What is DeepSeek V3.1 (Reasoning) best at?
Knowledge: 47.0 out of 100, place 157 of 268.
What is the context window of DeepSeek V3.1 (Reasoning)?
128K tokens.
Is DeepSeek V3.1 (Reasoning) open source?
Yes. Its weights are published.
When was DeepSeek V3.1 (Reasoning) released?
August 2025.

Badge

DeepSeek V3.1 (Reasoning) on Inference360: place 193 of 283, index 43.0
HTML
<a href="https://inference360.ai/models/deepseek-v3-1-reasoning"><img src="https://inference360.ai/badges/deepseek-v3-1-reasoning.svg" alt="DeepSeek V3.1 (Reasoning) on Inference360: place 193 of 283, index 43.0" height="20"></a>
Markdown
[![DeepSeek V3.1 (Reasoning) on Inference360: place 193 of 283, index 43.0](https://inference360.ai/badges/deepseek-v3-1-reasoning.svg)](https://inference360.ai/models/deepseek-v3-1-reasoning)

Compare with

K-Exaone42.3DeepSeek-R1-052844.6DeepSeek V3.137.9

More from DeepSeek

All models
DeepSeek V3.137.9DeepSeek V3.1 TerminusUnrankedDeepSeek V3.2 ExpUnrankedDeepSeek-R1-052844.6DeepSeekMath V2Unranked