Compare

GPT-5 Turbo vs DeepSeek V4

OpenAI vs DeepSeek. Specs, benchmarks, and real per-task cost — all in one page.

Verdict

GPT-5 Turbo leads on LMSYS ELO (1398 vs 1334). DeepSeek V4 is ~60.0× cheaper on a 3:1 input:output blend. DeepSeek V4 is open-weights, the other is proprietary.

OpenAI

GPT-5 Turbo

OpenAI’s flagship unified model. Handles text, vision, and audio natively. The generalist benchmark champion.

Native multimodalStrong math and scienceHuge ecosystem

DeepSeek

DeepSeek V4

The shock-the-market moment of 2026. Frontier coding quality at 1/50th the price of Opus.

Unbeatable $/qualityOpen weightsGreat at code

Pricing

Input / 1M$6.00$0.14

Output / 1M$24.00$0.28

Context400K128K

Max output32K16K

LicenseProprietaryOpen weights

Released2026-01-212026-02-20

Benchmarks

LMSYS ELO1398.01334.0

MMLU Pro91.285.4

HumanEval93.091.5

SWE-Bench65.854.1

MATH91.588.7

GPQA68.1—

MMMU78.8—

IFEval88.985.8

Per-task cost

Summarize a 1-hour meeting transcript$0.102$0.0022

Review a 500-line pull request$0.096$0.0017

Answer a customer support ticket$0.037$0.0006

Extract structured data from a resume$0.054$0.0010

Debug a stack trace with context$0.084$0.0014

Per-call cost using published token counts for each task. Real-world prompts vary.

Tool

Interactive compare →

Add up to four models, tweak tasks live.