Compare

DeepSeek R2 vs Claude Sonnet 4.6

DeepSeek vs Anthropic. 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.

DeepSeek

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

Frontier math at open-weights pricingLong CoT traces

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$0.55$3.00

Output / 1M$2.19$15.00

Context128K500K

Max output64K64K

LicenseOpen weightsProprietary

Released2026-01-152025-11-04

Benchmarks

LMSYS ELO—1368.0

MMLU Pro87.888.7

HumanEval90.192.1

SWE-Bench—64.5

MATH94.185.0

GPQA71.259.4

MMMU—74.1

IFEval—90.5

Per-task cost

Summarize a 1-hour meeting transcript$0.0093$0.052

Review a 500-line pull request$0.0088$0.054

Answer a customer support ticket$0.0034$0.021

Extract structured data from a resume$0.0049$0.030

Debug a stack trace with context$0.0077$0.048

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

Add up to four models, tweak tasks live.