Compare

Claude Sonnet 4.6 vs DeepSeek V4

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

Verdict

Claude Sonnet 4.6 leads on LMSYS ELO (1368 vs 1334). DeepSeek V4 is ~34.3× cheaper on a 3:1 input:output blend. DeepSeek V4 is open-weights, the other is proprietary.

Anthropic

Claude Sonnet 4.6

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 V4

The shock-the-market moment of 2026. Frontier coding quality at 1/50th the price of Opus.

Unbeatable $/qualityOpen weightsGreat at code

Pricing

Input / 1M$3.00$0.14

Output / 1M$15.00$0.28

Context500K128K

Max output64K16K

LicenseProprietaryOpen weights

Released2025-11-042026-02-20

Benchmarks

LMSYS ELO1368.01334.0

MMLU Pro88.785.4

HumanEval92.191.5

SWE-Bench64.554.1

MATH85.088.7

GPQA59.4—

MMMU74.1—

IFEval90.585.8

Per-task cost

Summarize a 1-hour meeting transcript$0.052$0.0022

Review a 500-line pull request$0.054$0.0017

Answer a customer support ticket$0.021$0.0006

Extract structured data from a resume$0.030$0.0010

Debug a stack trace with context$0.048$0.0014

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.