Compare

Claude Haiku 4.5 vs DeepSeek R2

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

Verdict

DeepSeek R2 is ~1.7× cheaper on a 3:1 input:output blend. DeepSeek R2 is open-weights, the other is proprietary.

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

DeepSeek

DeepSeek’s reasoning variant. Competes with o4 on math at a fraction of the cost.

Frontier math at open-weights pricingLong CoT traces

Pricing

Input / 1M$0.80$0.55

Output / 1M$4.00$2.19

Context200K128K

Max output16K64K

LicenseProprietaryOpen weights

Released2025-10-012026-01-15

Benchmarks

LMSYS ELO1305.0—

MMLU Pro81.487.8

HumanEval87.090.1

SWE-Bench49.2—

MATH76.894.1

GPQA—71.2

IFEval87.3—

Per-task cost

Summarize a 1-hour meeting transcript$0.014$0.0093

Review a 500-line pull request$0.014$0.0088

Answer a customer support ticket$0.0056$0.0034

Extract structured data from a resume$0.0080$0.0049

Debug a stack trace with context$0.013$0.0077

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

Add up to four models, tweak tasks live.