LIVE INTELLIGENCE · part of the Xentir AI Index

DeepSeek-V3 (Mar 2025) vs GPT-4.1

Name: DeepSeek-V3 (Mar 2025) vs GPT-4.1 — benchmark results
Creator: Epoch AI
License: https://creativecommons.org/licenses/by/4.0/

DeepSeek-V3 (Mar 2025) OPEN WEIGHTS · DeepSeek, released 2025-03-24
GPT-4.1 · OpenAI, released 2025-04-14

“Epoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks' [online resource].” · epoch.ai/benchmarks · Licensed CC BY 4.0 · data last verified 2026-09-09 · Methodology · Data Sources

GPT-4.1 leads on more of the 4 benchmarks where both are measured — DeepSeek-V3 (Mar 2025) on 1, GPT-4.1 on 3. The remaining 5 of 9 tracked benchmarks cannot be compared: 5 have a result for only one of the two models, 0 for neither. Nothing on this page is estimated to fill those in.

Head-to-head

Measured on 4 of 9 tracked benchmarks for both models — DeepSeek-V3 (Mar 2025) leads on 1, GPT-4.1 leads on 3.

Benchmark	DeepSeek-V3 (Mar 2025)	GPT-4.1	Leads
Epoch Capabilities Index (ECI)	135.9	136.8	GPT-4.1
SWE-bench Verified	Not measured	0.485	Comparison unavailable — DeepSeek-V3 (Mar 2025) has no published result
GPQA Diamond	0.676	0.669	DeepSeek-V3 (Mar 2025)
MATH Level 5	0.755	0.830	GPT-4.1
OTIS Mock AIME 2024-2025	0.378	0.383	GPT-4.1
FrontierMath	Not measured	0.055	Comparison unavailable — DeepSeek-V3 (Mar 2025) has no published result
FrontierMath Tier 4	Not measured	0.000	Comparison unavailable — DeepSeek-V3 (Mar 2025) has no published result
SimpleQA Verified	Not measured	0.311	Comparison unavailable — DeepSeek-V3 (Mar 2025) has no published result
Chess Puzzles	Not measured	0.060	Comparison unavailable — DeepSeek-V3 (Mar 2025) has no published result

Benchmark results measure specific tasks and should not be treated as a universal measure of model quality. A model with no row here for a benchmark has not been measured on it — that is left blank, never estimated.

What remains uncertain

DeepSeek-V3 (Mar 2025) has no published result on Chess Puzzles, FrontierMath, FrontierMath Tier 4, SWE-bench Verified, SimpleQA Verified.

A missing result means Xentir has not found one published by the source it cites, not that the model performs badly. See Data Sources.

Reference price

Per 1M tokens	DeepSeek-V3 (Mar 2025)	GPT-4.1
Output	—	$8.00

List price as published by the vendor. Xentir does not verify negotiated or enterprise rates, and price is not a measure of capability — see Data Sources.

Full profiles

DeepSeek-V3 (Mar 2025) → · GPT-4.1 → · the full AI Model Rankings table