Compare

Claude Haiku 4.5 vs Claude Sonnet 4.6

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

Verdict

Claude Sonnet 4.6 leads on LMSYS ELO (1368 vs 1305). Claude Haiku 4.5 is ~3.8× 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

Anthropic

The 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$0.80$3.00

Output / 1M$4.00$15.00

Context200K500K

Max output16K64K

LicenseProprietaryProprietary

Released2025-10-012025-11-04

Benchmarks

LMSYS ELO1305.01368.0

MMLU Pro81.488.7

HumanEval87.092.1

SWE-Bench49.264.5

MATH76.885.0

GPQA—59.4

MMMU—74.1

IFEval87.390.5

Per-task cost

Summarize a 1-hour meeting transcript$0.014$0.052

Review a 500-line pull request$0.014$0.054

Answer a customer support ticket$0.0056$0.021

Extract structured data from a resume$0.0080$0.030

Debug a stack trace with context$0.013$0.048

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

Add up to four models, tweak tasks live.