Compare

Claude Opus 4.7 vs DeepSeek R2

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

Verdict

DeepSeek R2 is ~31.3× cheaper on a 3:1 input:output blend. DeepSeek R2 is open-weights, the other is proprietary.

Anthropic

Anthropic’s frontier reasoning model. Leads SWE-Bench and holds top ranks on long-context coding and agentic workflows.

Best-in-class coding agent behaviorStrong instruction following1M-token context with high recallHigh reliability under tool use

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$15.00$0.55

Output / 1M$75.00$2.19

Context1.0M128K

Max output64K64K

LicenseProprietaryOpen weights

Released2026-03-122026-01-15

Benchmarks

LMSYS ELO1412.0—

MMLU Pro92.887.8

HumanEval94.490.1

SWE-Bench72.1—

MATH89.394.1

GPQA66.471.2

MMMU79.2—

IFEval92.1—

Per-task cost

Summarize a 1-hour meeting transcript$0.263$0.0093

Review a 500-line pull request$0.270$0.0088

Answer a customer support ticket$0.105$0.0034

Extract structured data from a resume$0.150$0.0049

Debug a stack trace with context$0.240$0.0077

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

Add up to four models, tweak tasks live.