DeepSeek vs Claude: prices, models and what each does better
The short answer
DeepSeek V4.1-Flash costs $0.15 per million input tokens and $0.60 per million output tokens off-peak. Claude Sonnet 5.5 costs $2.00 and $10.00: 13.3× as much on input and 16.7× on output. Against Claude Opus 5.5 and Claude Fable 5.2 the gap is larger still. The one Claude model in DeepSeek's price range is Claude Haiku 5.5.
Claude's strengths are agentic and coding work (Claude Code), careful writing, and the controls larger organisations ask for. DeepSeek's strengths are price, open weights and no long-context surcharge. Both read images and both handle about a million tokens of context.
Checked on 2026-11-19 against DeepSeek's pricing page and Anthropic's pricing documentation.
API prices per 1M tokens
| Model | Input | Cached input | Output |
|---|---|---|---|
| DeepSeek V4.1-Flash | $0.15 | $0.003 | $0.60 |
| DeepSeek V4-Pro | $0.66 | $0.022 | $1.98 |
| Claude Haiku 5.5 | $0.10 | $0.01 | $0.50 |
| Claude Sonnet 5.5 | $2.00 | $0.10 | $10.00 |
| Claude Opus 5.5 | $4.00 | $0.20 | $20.00 |
| Claude Fable 5.2 | $8.00 | $0.20 | $40.00 |
DeepSeek rates are off-peak; in DeepSeek's peak hours (01:00–04:00 and 06:00–10:00 UTC, Monday to Friday, excluding Chinese public holidays) they double. Claude Haiku 5.5 prices are for prompts up to 100K tokens. Cached input means a cache hit; writing to Claude's cache costs extra.
How many times more Claude costs
| Comparison | Input | Output |
|---|---|---|
| Claude Sonnet 5.5 vs DeepSeek V4.1-Flash | 13.3× | 16.7× |
| Claude Opus 5.5 vs DeepSeek V4-Pro | 6.1× | 10.1× |
| Claude Haiku 5.5 vs DeepSeek V4.1-Flash | 0.7× | 0.8× |
Note the last row: on short prompts Claude Haiku 5.5 is cheaper per token than DeepSeek V4.1-Flash. It is Anthropic's smallest model, so compare it on quality for your own task. Our DeepSeek pricing page has a calculator that compares a whole month of usage across providers.
The models side by side
DeepSeek offers two API models. DeepSeek-V4.1-Flash-1017 (deepseek-flash) is the default: 1M tokens of context, up to 384K output, and it understands images. DeepSeek-V4-Pro-0813 (deepseek-v4-pro) is the more expensive option for the hardest reasoning and is text only. The weights are open under the MIT licence. See all DeepSeek models.
Anthropic's current models are Claude Fable 5.2, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5. They take text and images and answer in text, with a 1M-token context window at the standard price (Haiku 5.5 charges more above 100K tokens). Around them Anthropic offers Claude Code for programming and apps for desktop and mobile.
Since 13 November 2026 DeepSeek's API also has web search, file input, OCR, speech-to-text and text-to-speech.
Claude plans vs DeepSeek's free app
| Claude plan | Price |
|---|---|
| Free | $0 |
| Pro | $20/mo, or $200 billed annually |
| Max | $100/mo — 5× Pro usage |
| Ultra | $200/mo |
US list prices as of 2026-11-19; check claude.com/pricing before you subscribe. DeepSeek's own app at chat.deepseek.com is free with an account and has no paid plan; see our login guide.
Which one to choose
Choose DeepSeek when cost per token decides, when you run a lot of text or image input through an API, or when you want open weights you can host yourself. Schedule batch work outside DeepSeek's peak hours to pay the lower rate.
Choose Claude for agentic and coding work with Claude Code, for writing where tone matters, or when your organisation needs Anthropic's enterprise controls. For cheap, short tasks, test Claude Haiku 5.5 against DeepSeek V4.1-Flash on your own prompts.
Why there is no benchmark table here
Many comparison pages print benchmark scores, language counts or speed figures that cannot be traced to either company or to a named leaderboard. Use Anthropic's own release posts, DeepSeek's model cards, and independent leaderboards such as Artificial Analysis, which date their results and publish their method. Or run your own prompts on both; DeepSeek's prices make that cheap.