Compare

Claude Sonnet 4.6 vs DeepSeek R2

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

Verdict

DeepSeek R2 is ~6.3× cheaper on a 3:1 input:output blend. DeepSeek R2 is open-weights, the other is proprietary.

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

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$3.00$0.55

Output / 1M$15.00$2.19

Context500K128K

Max output64K64K

LicenseProprietaryOpen weights

Released2025-11-042026-01-15

Benchmarks

LMSYS ELO1368.0—

MMLU Pro88.787.8

HumanEval92.190.1

SWE-Bench64.5—

MATH85.094.1

GPQA59.471.2

MMMU74.1—

IFEval90.5—

Per-task cost

Summarize a 1-hour meeting transcript$0.052$0.0093

Review a 500-line pull request$0.054$0.0088

Answer a customer support ticket$0.021$0.0034

Extract structured data from a resume$0.030$0.0049

Debug a stack trace with context$0.048$0.0077

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

Add up to four models, tweak tasks live.