GitHub[208K]試用 OpenCode

Compare AI models: DeepSeek V4 Flash vs Llama 4 Maverick 17B Instruct

  1. DeepSeek V4 FlashDeepSeek
  2. Llama 4 Maverick 17B InstructMetaHollow points use fallbacks · hover for details
ReasoningCodingCost efficiencyContext windowMultimodalTool use
Normalized model capability scores
ModelReasoningCodingCost efficiencyContext windowMultimodalTool use
DeepSeek V4 Flash25/10075/10058/10081/1000/1000/100
Llama 4 Maverick 17B Instruct0/100 — Reasoning not supported13/10065/10081/10033/10050/100 — Tool calling supported; no comparable benchmark
Overview
Author
Context length
1M
1M
Reasoning
True
False
Input modalities
Text
Text, Image
Output modalities
Text
Text
Providers
Pricing
Input
$0.30/ 1M
$0.27/ 1M
Output
$1.20/ 1M
$0.85/ 1M
Cached input
$0.006/ 1M
Unknown
MomentumLast 2 mo
Unique users
3.1M
No usage
Completed sessions
37,178,359
No usage
Token share
14.9%
No usage
Tokens
145T-37%
No usage
AUG 18OCT 12
No trend data
RetentionWeek 1
Returning users
69.3%272K user-weeks
Pending

Related comparisons. Other model pairs to check.