Compare

Claude Haiku 4.5 vs Grok 4

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

Verdict

Grok 4 leads on LMSYS ELO (1358 vs 1305). Claude Haiku 4.5 is ~4.7× cheaper on a 3:1 input:output blend.

Anthropic

Fast, cheap, and smart enough for most routing and extraction tasks. Great base model for subagents.

Sub-second latency10x cheaper than OpusStrong structured output

xAI

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

Live web/X accessPersonalityFast inference on Colossus

Pricing

Input / 1M$0.80$5.00

Output / 1M$4.00$15.00

Context200K256K

Max output16K32K

LicenseProprietaryProprietary

Released2025-10-012026-03-01

Benchmarks

LMSYS ELO1305.01358.0

MMLU Pro81.486.2

HumanEval87.088.4

SWE-Bench49.2—

MATH76.886.5

IFEval87.385.0

Per-task cost

Summarize a 1-hour meeting transcript$0.014$0.082

Review a 500-line pull request$0.014$0.070

Answer a customer support ticket$0.0056$0.027

Extract structured data from a resume$0.0080$0.040

Debug a stack trace with context$0.013$0.060

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

Add up to four models, tweak tasks live.