LocalMaxxing
Models
Leaderboard
Evals
Train
Rentals
API
Submit
Models
Leaderboard
Evals
Train
Rentals
API Docs
Language
English
简体中文
繁體中文
日本語
한국어
Total runs
Highest
Median
Lowest
Models
mlx-community
DeepSeek-V4-Flash-2bit-DQ
Total runs
Highest
Median
Lowest
Benchmarks
Evals
Benchmark Results
Submit benchmark
Any size
< 3B
3B
7B
13B
30B
70B
110B+
Advanced filters
#
Hardware
Engine · Quant
Ctx
tok/s out
prefill
tok/s total
TTFT ms
VRAM GB
DeepSeek-V4-Flash-2bit-DQ (284B) — 17.0 tok/s top speed