Anthropic: Claude Opus 4.5 vs DeepSeek: DeepSeek V3

DeepSeek: DeepSeek V3 is cheaper than Anthropic: Claude Opus 4.5 by about 97% on a blended (75% input / 25% output) basis. Anthropic: Claude Opus 4.5 charges $5.000000 / $25.000000 per 1M input/output tokens; DeepSeek: DeepSeek V3 charges $0.280000 / $0.420000.

FieldAnthropic: Claude Opus 4.5DeepSeek: DeepSeek V3
Input $/Mtok$5.000000$0.280000
Output $/Mtok$25.000000$0.420000
Cache read $/Mtok$0.500000$0.028000
Cache write $/Mtok$6.250000–
Input $/Mcharn/a0.091836
Output $/Mcharn/a0.137755
Context window200000163840
Tokenizerclaudedeepseek

Price history

Same workload, monthly cost

Illustrative monthly cost at 10M tokens/month, split per workload preset. Not a substitute for the calculator, which uses your own volume, cache and batch assumptions.

WorkloadAnthropic: Claude Opus 4.5DeepSeek: DeepSeek V3
Coding agent$60.00$2.87
RAG chat$80.00$3.01
Summarisation (batch)$70.00$2.94
Classification$54.00$2.83
Chat assistant$130.00$3.36
Content generation$210.00$3.92

Cite this page

Anthropic: Claude Opus 4.5 vs DeepSeek: DeepSeek V3: $5.000000/$25.000000 vs $0.280000/$0.420000 per 1M tokens, as of 2026-10-04 (Token Barometer, https://tokenbarometer.com/compare/anthropic_claude-opus-4-5-vs-deepseek_deepseek-chat).

Anthropic: Claude Opus 4.5 page | DeepSeek: DeepSeek V3 page | Open in calculator