Inference Index
Compare

Llama 4 Behemoth vs DeepSeek R2

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

Verdict

DeepSeek R2 is ~4.0× cheaper on a 3:1 input:output blend.

Meta
Llama 4 Behemoth

Meta’s open-weights flagship. 405B params, fully open license, runs on every major inference provider.

Best open-weights modelRuns anywhereTransparent training
DeepSeek
DeepSeek R2

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

Frontier math at open-weights pricingLong CoT traces
Pricing
Input / 1M$2.50$0.55
Output / 1M$8.00$2.19
Context256K128K
Max output16K64K
LicenseOpen weightsOpen weights
Released2025-12-102026-01-15
Benchmarks
LMSYS ELO1342.0
MMLU Pro87.187.8
HumanEval89.390.1
SWE-Bench52.4
MATH81.594.1
GPQA56.271.2
IFEval88.0
Per-task cost
Summarize a 1-hour meeting transcript$0.041$0.0093
Review a 500-line pull request$0.036$0.0088
Answer a customer support ticket$0.014$0.0034
Extract structured data from a resume$0.021$0.0049
Debug a stack trace with context$0.031$0.0077

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