AI Resources
Claude vs DeepSeek API pricing
Claude and DeepSeek API prices side by side: model tiers, current models, the cost of the same workload and what each API supports, read from the providers' own pricing pages.
Last verified: October 13, 2026
AnthropicOpus 5.5 and Fable 5.1 offer 1M context at standard token rates, with model-specific cache discounts.
DeepSeekV4.1-Flash provides 1M context and up to 384K output at $0.30/$1.20 peak or $0.15/$0.60 off-peak, with native vision on Flash.
Tier by tier
The closest model of each provider at four levels, shown as input / output price per million tokens.
| Tier | Claude | DeepSeek |
|---|---|---|
| Flagship | Fable 5.1$10/$50 | V4-Pro$1.32/$3.96 |
| Mid | Sonnet 5.5$2/$10 | V4.1-Flash$0.30/$1.20 |
| Budget | Haiku 4.5$1/$5 | V4.1-Flash$0.30/$1.20 |
| Reasoning | Opus 5.5$4/$20 | V4-Pro$1.32/$3.96 |
Cost of the same workload
Base token-cost estimate for 10 equal requests: 100K input + 10K output tokens each (1M + 100K total). Only models with documented limits covering each request are shown. DeepSeek uses peak rates; its cached row assumes all input hits an existing cache, excluding warm-up misses. Gemini 2.5 requires prior active use. Cache writes, storage, tools, media and service premiums are excluded.
| Claude | Cost |
|---|---|
| Claude Sonnet 5.5 | $3.00 |
| Claude Sonnet 5 | $3.00 |
| Claude Opus 5.5 | $6.00 |
| Claude Opus 4.6 | $7.50 |
| Claude Opus 5 | $7.50 |
| Claude Fable 5.1 | $15.00 |
| Claude Fable 5 | $15.00 |
| DeepSeek | Cost |
|---|---|
| DeepSeek V4.1-Flashcached input | $0.13 |
| DeepSeek V4.1-Flash | $0.42 |
| DeepSeek V4-Pro | $1.72 |
Anthropic models and prices
| Model | API ID | Input | Output | Batch in/out | Cache hit | Context | Max out |
|---|---|---|---|---|---|---|---|
| Fable 5.1 | claude-fable-5-1 | $10.00 | $50.00 | $5.00 / $25.00 | $0.25 | 1M | 128K |
| Opus 5.5 | claude-opus-5-5 | $4.00 | $20.00 | $2.00 / $10.00 | $0.20 | 1M | 128K |
| Mythos 5.1 | claude-mythos-5-1 | $10.00 | $50.00 | $5.00 / $25.00 | $0.25 | 1M | 128K |
| Mythos 5 | claude-mythos-5 | $10.00 | $50.00 | $5.00 / $25.00 | $1.00 | 1M | 128K |
| Sonnet 5.5 | claude-sonnet-5-5 | $2.00 | $10.00 | $1.00 / $5.00 | $0.10 | 1M | 128K |
| Haiku 5.5 | claude-haiku-5-5 | $0.10 / $0.50 | $0.50 / $2.50 | $0.05 / $0.25; $0.25 / $1.25 | $0.01 / $0.05 | 1M | 128K |
Fable 5.1 and Mythos 5.1 cache reads cost 2.5% of base input; Opus 5.5 and Sonnet 5.5 reads cost 5%; other listed models use 10%. Five-minute writes cost 1.25x input and one-hour writes 2x, without a separate storage fee. Batch discounts input and output by 50%. Fable 5.1, Opus 5.5 and Sonnet 5.5 provide 1M context at standard rates. Haiku 5.5 has 1M context but higher rates above 100K prompt tokens, including cached tokens. Its directory cells show up-to-100K then above-100K rates; each batch pair is input/output. Output limits shown are standard limits, excluding Batch betas. Opus 4.1 retired on August 5, 2026. Opus 5.5, Sonnet 5.5 and Haiku 5.5 additionally support 300K maximum output on the Message Batches API with the output-300k-2026-03-24 beta header. Sonnet 4.5 was deprecated on 2026-09-30 and is scheduled to retire on 2026-11-30; Anthropic recommends claude-sonnet-5-5.
DeepSeek models and prices
| Model | API ID | Input (miss) | Input (hit) | Output | Context | Max out |
|---|---|---|---|---|---|---|
| V4.1-Flash | deepseek-flash | $0.30 | $0.006 | $1.20 | 1M | 384K |
| V4-Pro | deepseek-v4-pro | $1.32 | $0.044 | $3.96 | 1M | 384K |
Selected current API models and historical aliases. Directory, charts, simulation and KPI use peak rates: V4.1-Flash $0.30/$1.20 and V4-Pro $1.32/$3.96 input/output. Peak is 01:00-04:00 and 06:00-10:00 UTC Monday-Friday, excluding Chinese public holidays; all other hours, weekends and Chinese public holidays are off-peak. Peak cached input is $0.006 on Flash and $0.044 on Pro. Automatic caching is best-effort. The fully cached simulation assumes every input token hits an existing cache and excludes warm-up misses. Both models publish 1M context and 384K maximum output. Concurrency is 2,500 on Flash and 500 on Pro. Capabilities marked not documented refer only to the reviewed API sources, not a claim of provider-wide absence.
What each API supports
| Capability | Claude | DeepSeek |
|---|---|---|
| Vision (image in) | Yes | V4.1-Flash only; not V4-Pro |
| Video input | Not documented | Not documented |
| Audio input | Not documented | Not documented |
| PDF input | PDF text and page images | Not documented |
| Image generation | Not documented | Not documented |
| Video generation | Not documented | Not documented |
| Text to speech | Not documented | Not documented |
| Embeddings | Not documented | Not documented |
| Web search | $10/1K | Not documented |
| Tool / function calling | Yes | Yes |
| Structured JSON | Schema enforced | JSON output; validate the result |
| Reasoning mode | Adaptive thinking | Thinking mode |
| Code fill-in-the-middle | Not documented | Non-thinking only; 4K max output (Beta) |
| Prompt caching | 90–97.5% off cache reads | Automatic; Flash 98%, Pro about 97% off |
| Batch API | Yes, 50% off | Not documented |
| Long-context surcharge | Haiku 5.5: higher rates above 100K prompt tokens | None |
| Free tier | Trial credits | Not documented |
| Max context | 1M | 1M |
| Max output | 128K standard; 300K on selected Batch betas | 384K |
| OpenAI SDK compatible | Testing layer; no prompt caching or strict schemas | Plus Anthropic API |
| Data sovereignty | US residency (1.1x) | Not documented |
Sources
Every figure on this page was read from the provider's own pricing documentation on the verification date shown above.