About
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...
Related Models
DeepSeek V4 Pro (Reasoning, High Effort)2026-04-24DeepSeek: DeepSeek V4 Flash2026-04-24DeepSeek V4 Flash (Reasoning, High Effort)2026-04-24DeepSeek V4 Pro (Non-reasoning)2026-04-24DeepSeek V4 Flash (Non-reasoning)2026-04-24DeepSeek: DeepSeek V4 Flash (free)2026-04-24DeepSeek V3.2 (Reasoning)2025-12-01DeepSeek: DeepSeek V3.22025-12-01
Pricing
Input
$0.43
per 1M tokens
Output
$0.87
per 1M tokens
Blended
$0.54
per 1M tokens
Cheaper than 50% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.54
Monthly
$16.31
vs. Similar Models
OpenAI: GPT-5.3-CodexQ:0.0
$4.81+785%
MiniMax: MiniMax M3Q:+0.1
$0.52-3%
Claude Opus 4.6 (Adaptive Reasoning, Max Effort)Q:-0.6
$10.00+1739%
GPT-5.5 (low)Q:-0.8
$11.25+1969%
Performance
84
tokens/sec
Faster than 43% of models
1.18
seconds
Faster than 45% of models
53.49
seconds
Faster than 6% of models
Market Median
94 tok/s
11% slower
Median TTFT
1.11s
6% slower
Throughput/Dollar
154
tok/s per $/1M
Speed Comparison
DeepSeek V4 Pro (Non-reasoning)
84 tok/s+0%
Qwen: Qwen3.5-27B
84 tok/s+0%
GPT-5.5 (high)
83 tok/s-1%
Context Window
1.0M
tokens
Larger than 90% of models
Max Output
384K
tokens
37% of context
Benchmarks
MMLU-ProNot evaluated
GPQA Diamond
88.8%
HLE
35.9%
LiveCodeBenchNot evaluated
SciCode
50.0%
TerminalBench Hard
46.2%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025Not evaluated
IFBench
76.5%
Long Context Recall
66.3%
Tau2
96.2%
Market AverageTop Score