Compare

Claude Haiku 4.5 vs Qwen 3 Max

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

Verdict

Claude Haiku 4.5 leads on LMSYS ELO (1305 vs 1295). Claude Haiku 4.5 is ~1.8× cheaper on a 3:1 input:output blend. Qwen 3 Max is open-weights, the other is proprietary.

Anthropic

Claude Haiku 4.5

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

Sub-second latency10x cheaper than OpusStrong structured output

Qwen

Qwen 3 Max

Alibaba’s flagship. Strongest non-English coverage of any open model; dominant in Asian markets.

MultilingualOpen weightsCompetitive coding

Pricing

Input / 1M$0.80$1.60

Output / 1M$4.00$6.40

Context200K128K

Max output16K16K

LicenseProprietaryOpen weights

Released2025-10-012025-11-20

Benchmarks

LMSYS ELO1305.01295.0

MMLU Pro81.484.0

HumanEval87.087.2

SWE-Bench49.2—

MATH76.885.1

MMMU—71.8

IFEval87.385.5

Per-task cost

Summarize a 1-hour meeting transcript$0.014$0.027

Review a 500-line pull request$0.014$0.026

Answer a customer support ticket$0.0056$0.0099

Extract structured data from a resume$0.0080$0.014

Debug a stack trace with context$0.013$0.022

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.