Qwen3.8 DFlash2 benchmark against Optimized Speed and Quality MTP - #304

Open

davidtai wants to merge 16 commits into

youssofal:mainfrom

davidtai:perf/qwen38-dflash2

Commits on Aug 20, 2026

docs: design Qwen3.8 DFlash2 benchmark
davidtai
committed
75c81b9
docs: split DFlash2 depth sweep from optimization
davidtai
committed
4623a99
Define DFlash2 depth sweep contract
davidtai
committed
6e2af44
Upgrade dflash-mlx to upstream head
davidtai
committed
d6f8da8
docs: record DFlash dependency constraint
davidtai
committed
62281c3
Connect DFlash2 to an MTPLX target
davidtai
committed
4d3d03a
docs: use Qwen3.8 DFlash2 block width
davidtai
committed
520ac13
Add greedy DFlash2 depth sweep
davidtai
committed
5fdad7d
Wire Qwen3.8 DFlash2 benchmark channel
davidtai
committed
dbfce28
docs: record DFlash2 verification
davidtai
committed
2949145
Pin DFlash cache ownership fix
davidtai
committed
e68f1e9
Document DFlash cache ownership fix
davidtai
committed
91fb957
Standardize greedy DFlash benchmark sampling
davidtai
committed
f9b7177
Classify DFlash benchmark token drift
davidtai
committed
9d87e50
Track DFlash prefill and hit-rate metrics
davidtai
committed
c3487dc

Commits on Aug 21, 2026

Benchmark DFlash2 against Qwen3.8 Quality
davidtai
committed
3a7c9f1