Compare

Grok 4 vs Llama 4 Behemoth

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

Verdict

Grok 4 leads on LMSYS ELO (1358 vs 1342). Llama 4 Behemoth is ~1.9× cheaper on a 3:1 input:output blend. Llama 4 Behemoth is open-weights, the other is proprietary.

xAI

Grok 4

xAI’s latest. Real-time X search integration, strong on current events and meme-literate tasks.

Live web/X accessPersonalityFast inference on Colossus

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

Pricing

Input / 1M$5.00$2.50

Output / 1M$15.00$8.00

Context256K256K

Max output32K16K

LicenseProprietaryOpen weights

Released2026-03-012025-12-10

Benchmarks

LMSYS ELO1358.01342.0

MMLU Pro86.287.1

HumanEval88.489.3

SWE-Bench—52.4

MATH86.581.5

GPQA—56.2

IFEval85.088.0

Per-task cost

Summarize a 1-hour meeting transcript$0.082$0.041

Review a 500-line pull request$0.070$0.036

Answer a customer support ticket$0.027$0.014

Extract structured data from a resume$0.040$0.021

Debug a stack trace with context$0.060$0.031

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.