deepseek-v4-flash vs gemini-3.1-flash-lite-preview
Pricing, Performance & Features Comparison
Context Length1M
Reasoning
Providers1
ReleasedApr 2026
Knowledge Cutoff–
LicenseMIT License
Mixture-of-Experts model with 284B total parameters and 13B activated per token. Features hybrid attention architecture for efficient 1M context processing.
Input$0.14
Output$0.28
Latency (p50)2.3s
Output Limit384K
Function Calling
JSON Mode
-
InputText
OutputText
in$0.14out$0.28cache$0.028write$0.14
Latency (24h)
Success Rate (24h)
Gemini 2.0 Flash-Lite is Google's most cost-efficient Gemini 2.0 model, optimized for cost efficiency and low latency. It is a multimodal model supporting inputs including text, code, images, audio, and video, with text-only output. It features a maximum input capacity of 1,048,576 tokens and is cost-optimized for large-scale text output use cases.
Input$0.25
Output$1.5
Latency (p50)2.4s
Output Limit8K
Function Calling
JSON Mode
InputText, Image, Audio, Video
OutputText
in$0.25out$1.5cache$0.025–