DeepSeek: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by DeepSeek for your use-case.
Most Intelligent
DeepSeek V4.1 Flash 1017 (max)DeepSeek V4 Pro 0813 (max)DeepSeek V4 Pro (max)DeepSeek V4 Pro (high)DeepSeek V4 Pro (non-reasoning)Intelligence Index
Total 6 models
Fastest
DeepSeek V4.1 Flash 1017 (max)DeepSeek V4 Pro (max)DeepSeek V4 Pro (high)DeepSeek V4 Pro 0813 (max)DeepSeek V4 Pro 0813 (non-reasoning)Output speed
Total 6 models
Lowest Price
DeepSeek V4 Pro (max)DeepSeek V4 Pro (high)DeepSeek V4 Pro (non-reasoning)DeepSeek V4.1 Flash 1017 (max)DeepSeek V4 Pro 0813 (max)Blended price (per 1M tokens)
Total 6 models
DeepSeek offers 6 models, each with different intelligence, performance, and pricing characteristics. Below is a comparison of the key metrics across models.
- For intelligence, the top models on DeepSeek are DeepSeek V4.1 Flash 1017 (max) (45), DeepSeek V4 Pro 0813 (max) (36), and DeepSeek V4 Pro (max) (30).
- For output speed, the fastest models are DeepSeek V4.1 Flash 1017 (max) (195 t/s), DeepSeek V4 Pro (max) (102 t/s), and DeepSeek V4 Pro (high) (101 t/s). Speed varies significantly across models, with a 103% difference between the fastest and slowest.
- For latency, DeepSeek V4 Pro (non-reasoning) (1.61s), DeepSeek V4 Pro 0813 (non-reasoning) (1.71s), and DeepSeek V4.1 Flash 1017 (max) (11.39s) offer the lowest time to first answer token.
- For pricing, DeepSeek V4 Pro (max) ($0.18), DeepSeek V4 Pro (high) ($0.18), and DeepSeek V4 Pro (non-reasoning) ($0.18) offer the lowest blended prices per 1M tokens. Prices vary up to 3.9x across models.
- For context window size, DeepSeek V4 Pro 0813 (max) (1M), DeepSeek V4.1 Flash 1017 (max) (1M), and DeepSeek V4 Pro (max) (1M) support the largest context windows on DeepSeek.
- DeepSeek V4.1 Flash 1017 (max) offers the best combination of intelligence and speed. For cost optimization, DeepSeek V4 Pro (max) provides the most competitive pricing.
Intelligence Evaluations
Artificial Analysis Intelligence Index
Evaluations
Agentic knowledge work, Elo
Knowledge
1 - hallucination rate
Agentic SaaS workflows
Agentic real-world work tasks, Elo
Reasoning & knowledge
Professional document reasoning, All-pass
Agentic coding & terminal use
Long context reasoning
Coding
Physics reasoning
Agentic scientific research workflows in a terminal
Agentic software engineering in open-source repositories
Quantitative analysis on spreadsheets & documents
Legal agentic work, Hallucination-Gated All-Pass Rate
Agentic business operations
Agentic interactive 3D experiences, Elo
Kubernetes incident root-cause analysis
Medical long context reasoning
Visual reasoning
Rich, varied, well-crafted language
Well-structured, scannable, answer-first
Truthful, acknowledges uncertainty, non-sycophantic
Use of filler and generic language
Intelligence Index vs. Price
Context Window
Context Window
Pricing
Intelligence Index vs. Price
Performance Summary
Output Speed vs. Price
Speed
Measured by Output Speed (tokens per second)
Output Speed
Latency
Measured by Time (seconds) to First Token
Latency: Time To First Answer Token
Cache Behaviour
Cache Hit Rate
Cost per Task vs. Cache Hit Rate
End-to-End Response Time
Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed
End-to-End Response Time vs. Price
Further Analysis | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
DeepSeek V4.1 Flash 1017 (max) | 1M | Open | 45 | $0.26 | 201 | 1.10 | 13.52 | 9.94 | |||
DeepSeek V4 Pro 0813 (max) | 1.05M | Open | 36 | $0.67 | 93 | 1.65 | 28.67 | 21.61 | |||
DeepSeek V4 Pro (max) | 1M | Open | 30 | $0.12 | 100 | 1.83 | 50.61 | 43.78 | |||
DeepSeek V4 Pro (high) | 1M | Open | 30* | -- | 93 | 1.70 | 28.37 | 21.32 | |||
DeepSeek V3.2 | 128k | Open | 21* | -- | -- | -- | -- | -- | |||
DeepSeek V4 Pro (non-reasoning) | 1M | Open | 21* | -- | 88 | 1.65 | 7.36 | -- | |||
DeepSeek V4 Pro 0813 (non-reasoning) | 1M | Open | 20 | $0.36 | 96 | 1.69 | 6.90 | -- | |||
DeepSeek V3.2 Exp | 128k | Open | 17* | -- | -- | -- | -- | -- | |||
DeepSeek V3.2 (non-reasoning) | 128k | Open | 16* | -- | -- | -- | -- | -- | |||
DeepSeek V3.2 Exp (non-reasoning) | 128k | Open | 14* | -- | -- | -- | -- | -- | |||
Key definitions
Frequently Asked Questions
Common questions about DeepSeek
DeepSeek offers 6 models that we track: DeepSeek V4.1 Flash 1017 (max), DeepSeek V4 Pro 0813 (max), DeepSeek V4 Pro (max), DeepSeek V4 Pro (high), DeepSeek V4 Pro (non-reasoning), and DeepSeek V4 Pro 0813 (non-reasoning).
The most intelligent model available on DeepSeek is DeepSeek V4.1 Flash 1017 (max) with an Intelligence Index score of 45.
The fastest model on DeepSeek by output speed is DeepSeek V4.1 Flash 1017 (max) at 195.2 tokens per second.
The model with the lowest time to first answer token on DeepSeek is DeepSeek V4 Pro (non-reasoning) at 1.61s. Lower latency means faster initial response time.
The most affordable model on DeepSeek by blended price is DeepSeek V4 Pro (max) at $0.18 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on DeepSeek vary up to 4x across models, from $0.18 per 1M tokens for DeepSeek V4 Pro (max) to $0.69 per 1M tokens for DeepSeek V4 Pro 0813 (non-reasoning).
Yes, DeepSeek offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
Yes, all 6 models on DeepSeek support JSON mode for structured output.
Yes, all 6 models on DeepSeek support function calling (tool use).
Yes, DeepSeek offers 4 reasoning models: DeepSeek V4.1 Flash 1017 (max), DeepSeek V4 Pro 0813 (max), DeepSeek V4 Pro (max), and DeepSeek V4 Pro (high). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, all 6 models on DeepSeek are open weight models.
Yes, provider performance can vary over time due to infrastructure changes, load balancing, and updates. We continuously benchmark all providers and display historical performance trends in the "Over Time" charts.
When choosing a model on DeepSeek, consider: intelligence (for quality-sensitive tasks), output speed (for throughput-intensive tasks), latency (for interactive applications requiring quick first responses), pricing (for cost-sensitive workloads), and features like context window size, JSON mode, or function calling support.