Market preview
DeepSeek: DeepSeek Pro Latest+27%DeepSeek V4 Flash Latest+16%DeepSeek: DeepSeek V4 Flash 0731+16%@cf/ai4bharat/indictrans2-en-indic-1B$0.342 / 1MV100$0.02/hr
DeepSeek · Tools · Structured output · Reasoning controls · Sampling controls · Output limits · Released Aug 21, 2026
Provider pricing
Published rates for providers serving this model. Open any source to see where and when the price was checked.
| Provider | Input | Output | Cached input | Batch input | Evidence |
|---|---|---|---|---|---|
 OpenRouter openrouter.ai ↗ | $0.216/ 1M | $0.647/ 1M | $0.0069 | Not published | Sourcehttps://openrouter.ai/api/v1/models See where this value came from and when it was checked.
|
 DeepInfradeepinfra.com ↗ | $0.44/ 1M | $1.32/ 1M | $0.014 | Not published | Sourcehttps://api.deepinfra.com/v1/openai/models See where this value came from and when it was checked.
|
 Fireworks AIfireworks.ai ↗ | $0.22/ 1M | $0.66/ 1M | $0.007 | Not published | Sourcehttps://app.fireworks.ai/api/models See where this value came from and when it was checked.
|
| deepseek.com ↗ | $0.44/ 1M | $1.32/ 1M | $0.014 | Not published | Sourcehttps://api-docs.deepseek.com/quick_start/pricing See where this value came from and when it was checked.
|
Performance
Benchmark results collected from published sources. Scores from different benchmark versions or test setups may not be directly comparable.
| Benchmark | Category | Score | Origin | Evidence |
|---|---|---|---|---|
| GPQA Diamond | reasoning | 90.00 | Public benchmark | SourceGPQA Diamond via OpenRouter — measured on deepseek-v4-flash-0731 See where this value came from and when it was checked.
|
| TAU-Bench | reasoning | 77.80 | Public benchmark | SourceTAU-Bench via OpenRouter — measured on deepseek-v4-flash-0731 See where this value came from and when it was checked.
|
| Artificial Analysis Intelligence | reasoning | 51.80 | Public benchmark | SourceArtificial Analysis Intelligence via OpenRouter — measured on deepseek-v4-flash-0731 See where this value came from and when it was checked.
|
| Artificial Analysis Coding | reasoning | 69.10 | Public benchmark | SourceArtificial Analysis Coding via OpenRouter — measured on deepseek-v4-flash-0731 See where this value came from and when it was checked.
|
| Artificial Analysis Agentic | reasoning | 48.40 | Public benchmark | SourceArtificial Analysis Agentic via OpenRouter — measured on deepseek-v4-flash-0731 See where this value came from and when it was checked.
|
History
Recorded changes in pricing, availability and other tracked model data.