Compare
Llama 4 Behemoth vs GPT-5 Mini
Meta vs OpenAI. Specs, benchmarks, and real per-task cost — all in one page.
Verdict
Llama 4 Behemoth leads on LMSYS ELO (1342 vs 1288). GPT-5 Mini is ~4.4× cheaper on a 3:1 input:output blend. Llama 4 Behemoth is open-weights, the other is proprietary.
Meta
Llama 4 BehemothMeta’s open-weights flagship. 405B params, fully open license, runs on every major inference provider.
Best open-weights modelRuns anywhereTransparent training
OpenAI
GPT-5 MiniGPT-5’s smaller, cheaper sibling. Optimized for chat and lightweight agent tasks.
Very low latencyCheap enough for mass consumer apps
Pricing
Input / 1M$2.50$0.50
Output / 1M$8.00$2.00
Context256K256K
Max output16K16K
LicenseOpen weightsProprietary
Released2025-12-102026-01-21
Benchmarks
LMSYS ELO1342.01288.0
MMLU Pro87.178.5
HumanEval89.385.5
SWE-Bench52.4—
MATH81.579.8
GPQA56.2—
IFEval88.085.4
Per-task cost
Summarize a 1-hour meeting transcript$0.041$0.0085
Review a 500-line pull request$0.036$0.0080
Answer a customer support ticket$0.014$0.0031
Extract structured data from a resume$0.021$0.0045
Debug a stack trace with context$0.031$0.0070
Per-call cost using published token counts for each task. Real-world prompts vary.