About
DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and...
Related Models
DeepSeek: DeepSeek V4 Pro2026-04-24DeepSeek V4 Pro (Reasoning, High Effort)2026-04-24DeepSeek 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
$0.09
per 1M tokens
Output
$0.18
per 1M tokens
Blended
$0.11
per 1M tokens
Cheaper than 82% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.11
Monthly
$3.38
vs. Similar Models
MiMo-V2-ProQ:0.0
$1.50+1233%
Z.ai: GLM 5.1Q:-0.1
$1.50+1238%
MiMo-V2.5Q:-0.2
$0.17+56%
OpenAI: GPT-5.2-CodexQ:-0.2
$4.81+4178%
Performance
112
tokens/sec
Faster than 60% of models
0.99
seconds
Faster than 57% of models
51.22
seconds
Faster than 6% of models
Market Median
95 tok/s
18% faster
Median TTFT
1.11s
11% faster
Throughput/Dollar
993
tok/s per $/1M
Speed Comparison
GPT-5.4 (low)
111 tok/s-0%
Qwen3 30B A3B (Non-reasoning)
112 tok/s+0%
Nex AGI: Nex-N2-Pro
111 tok/s-1%
Context Window
1.0M
tokens
Larger than 90% of models
Max Output
66K
tokens
6% of context
Benchmarks
MMLU-ProNot evaluated
GPQA Diamond
89.4%
HLE
32.1%
LiveCodeBenchNot evaluated
SciCode
44.9%
TerminalBench Hard
35.6%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025Not evaluated
IFBench
79.2%
Long Context Recall
63.0%
Tau2
95.0%
Market AverageTop Score