Compare

Llama 4 Scout vs DeepSeek V4

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

Verdict

DeepSeek V4 leads on LMSYS ELO (1334 vs 1265). DeepSeek V4 is ~2.6× cheaper on a 3:1 input:output blend.

Meta

The 70B workhorse open-weight. Fits on a single H100, beloved by self-hosters.

Single-GPU deployableDirt cheap on aggregatorsStrong fine-tuning base

DeepSeek

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$0.35$0.14

Output / 1M$0.80$0.28

Context128K128K

Max output8K16K

LicenseOpen weightsOpen weights

Released2025-12-102026-02-20

Benchmarks

LMSYS ELO1265.01334.0

MMLU Pro77.285.4

HumanEval82.491.5

SWE-Bench—54.1

MATH72.588.7

IFEval84.185.8

Per-task cost

Summarize a 1-hour meeting transcript$0.0057$0.0022

Review a 500-line pull request$0.0044$0.0017

Answer a customer support ticket$0.0017$0.0006

Extract structured data from a resume$0.0026$0.0010

Debug a stack trace with context$0.0037$0.0014

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

Add up to four models, tweak tasks live.