Compare
Llama 4 Behemoth vs Claude Sonnet 4.6
Meta vs Anthropic. Specs, benchmarks, and real per-task cost — all in one page.
Verdict
Claude Sonnet 4.6 leads on LMSYS ELO (1368 vs 1342). Llama 4 Behemoth is ~1.5× 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
Anthropic
Claude Sonnet 4.6The workhorse. Near-Opus quality at 1/5 the cost. The default choice for production code and agent workloads.
Excellent cost/quality ratioFast inferenceGreat for Cursor/Claude Code
Pricing
Input / 1M$2.50$3.00
Output / 1M$8.00$15.00
Context256K500K
Max output16K64K
LicenseOpen weightsProprietary
Released2025-12-102025-11-04
Benchmarks
LMSYS ELO1342.01368.0
MMLU Pro87.188.7
HumanEval89.392.1
SWE-Bench52.464.5
MATH81.585.0
GPQA56.259.4
MMMU—74.1
IFEval88.090.5
Per-task cost
Summarize a 1-hour meeting transcript$0.041$0.052
Review a 500-line pull request$0.036$0.054
Answer a customer support ticket$0.014$0.021
Extract structured data from a resume$0.021$0.030
Debug a stack trace with context$0.031$0.048
Per-call cost using published token counts for each task. Real-world prompts vary.