Compare

DeepSeek R2 vs Llama 4 Scout

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

Verdict

Llama 4 Scout is ~2.1× cheaper on a 3:1 input:output blend.

DeepSeek

DeepSeek’s reasoning variant. Competes with o4 on math at a fraction of the cost.

Frontier math at open-weights pricingLong CoT traces

Meta

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

Single-GPU deployableDirt cheap on aggregatorsStrong fine-tuning base

Pricing

Input / 1M$0.55$0.35

Output / 1M$2.19$0.80

Context128K128K

Max output64K8K

LicenseOpen weightsOpen weights

Released2026-01-152025-12-10

Benchmarks

LMSYS ELO—1265.0

MMLU Pro87.877.2

HumanEval90.182.4

MATH94.172.5

GPQA71.2—

IFEval—84.1

Per-task cost

Summarize a 1-hour meeting transcript$0.0093$0.0057

Review a 500-line pull request$0.0088$0.0044

Answer a customer support ticket$0.0034$0.0017

Extract structured data from a resume$0.0049$0.0026

Debug a stack trace with context$0.0077$0.0037

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

Add up to four models, tweak tasks live.