---
title: "Deepseek V3 671B Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for Deepseek V3."
source: "https://rankedagi.com/models/deepseek-3-old"
---

# Deepseek V3

| Field | Value |
| --- | --- |
| Organization | DeepSeek |
| License | Open Source |
| Size | 671B |
| Version | Old |
| Released | 2024-12-26 |
| Context window | 128K |
| Knowledge cutoff | 2024-07-01 |
| Input cost per million tokens | $0.07 |
| Output cost per million tokens | $1.1 |
| Cached input cost per million tokens | $0.27 |
| Model ID | deepseek-chat |
| Last updated | 2026-05-04T01:34:22.840Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 40.0% | RankedAGI Coding Score |  |
| SWEBench Verified | coding | 42.0% | Agentic Coding | https://www.swebench.com/index.html |
| RankedAGI Agentic | Agents | 34.5% | RankedAGI Agentic Score |  |
| LiveCodeBench v5 | coding | 40.5% | LiveCodeBench - Evaluation of Large Language Models for Code |  |
| ChatArena (LMSYS) | coding | 1279 | ChatArena (LMSYS) Coding ELO Score with Style Control |  |
| Codeforces ELO Score | coding | 1134 | Codeforces ELO Score |  |
| Aider Polyglot | coding | 48.4% | Aider Polyglot Code Completion Benchmark |  |
| LiveBench Coding Score (old) | coding | 61.8% | LiveBench Coding Score old version |  |
| RankedAGI Reasoning | reasoning | 38.4% | RankedAGI Reasoning Score |  |
| GPQA Diamond | reasoning | 59.1% | Generalized Prefix Question Answering Score (Reasoning) PhD Level Reasoning |  |
| Text Arena | general | 1358 | ChatArena (LMSYS) ELO Score | https://arena.ai/leaderboard/text |
| AIME 2025 | math | 25.0% | AIME 2025 Competition Math |  |
| HumanEval | coding | 82.6% | HumanEval coding benchmark |  |
| HumanEval+ | coding | 86.6% | HumanEval+ benchmark |  |
| NYT Connections Extended | reasoning | 15.1% | NYT Connections Extended Version | https://github.com/lechmazur/nyt-connections/ |
| MMLU Pro | knowledge | 75.9% | Graduate-level visual problem-solving |  |
| MMLU | knowledge | 88.5% | Image Reasoning |  |
| Hughes Hallucination Evaluation | safety | 3.9% | Computed using Vectara's Hallucination Evaluation Model, also known as HHEM. This evaluates how often an LLM introduces hallucinations when summarizing a document. | https://github.com/vectara/hallucination-leaderboard |
| LiveBench Average | general | 60.5% | LiveBench Average Score old version |  |
| Instruction Following Evaluation | instruction-following | 86.1% | Instruction Following Evaluation |  |
| LiveBench Data Analysis | general | 60.9% | LiveBench Data Analysis Score | https://livebench.ai/ |
| LiveBench Language | general | 47.5% | LiveBench Language Score | https://livebench.ai/ |
| RankedAGI Math | math | 44.5% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 42.1% | Overall RankedAGI score |  |