Compare

Claude Haiku 4.5 vs DeepSeek V4

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

Verdict

DeepSeek V4 leads on LMSYS ELO (1334 vs 1305). DeepSeek V4 is ~9.1× cheaper on a 3:1 input:output blend. DeepSeek V4 is open-weights, the other is proprietary.

Anthropic

Claude Haiku 4.5

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 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$0.80$0.14

Output / 1M$4.00$0.28

Context200K128K

Max output16K16K

LicenseProprietaryOpen weights

Released2025-10-012026-02-20

Benchmarks

LMSYS ELO1305.01334.0

MMLU Pro81.485.4

HumanEval87.091.5

SWE-Bench49.254.1

MATH76.888.7

IFEval87.385.8

Per-task cost

Summarize a 1-hour meeting transcript$0.014$0.0022

Review a 500-line pull request$0.014$0.0017

Answer a customer support ticket$0.0056$0.0006

Extract structured data from a resume$0.0080$0.0010

Debug a stack trace with context$0.013$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.