List your company

Code Generation on HumanEval (TPS, Speedup, Performance Score)

351.07TPS (Tokens/s)EVICT
-2.7989.0775180.945272.8125Mar 10, 2026Mar 18, 2026Mar 27, 2026Apr 5, 2026Apr 13, 2026Apr 22, 2026May 1, 2026

Evaluation results

MethodLinks
EVICTTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05351.07——
EVICTTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05335.78——
ES-dLLMBase Model=Dream-7B-Base, Few-shot=02026.03305.376.938.41
DualCacheBase Model=Dream-7B-Base, Few-shot=02026.03257.425.840.24
EAGLE-3Temperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05254.55——
DDDTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05253.06——
EAGLE-3Temperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05249.55——
DDDTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=4, steps=4, topk=8, draft_tokens=322026.05244.41——
VanillaTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05174.51——
VanillaTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05169.86——
LookaheadTemperature=0, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05130.89——
LookaheadTemperature=1, Backbone=Ling-flash-2.0, Batch Size=1, TP (Tensor Parallelism)=42026.05128.99——
STDecBase Model=Dream-7B-Instruct, Decoding Strategy=STDec2026.0464.55.4660.37
LocalLeapBase Model=Dream-7B-Instruct, Decoding Strategy=LocalLeap2026.0453.84.5558.54
STDecBackbone=LLaDA-8B-Instruct2026.0452.924.8948.78
LLaDA + LocalLeapBackbone=LLaDA-8B-Instruct2026.0450.334.6546.34
Fast-dLLM (Parallel)Base Model=Dream-7B-Instruct, Decoding Strategy=Fast-dLLM (Parallel)2026.0446.873.9762.2
DreamBase Model=Dream-7B-Base, Few-shot=02026.0344.26142.07
LLaDA + Fast-dLLM (Parallel)Backbone=LLaDA-8B-Instruct2026.0438.493.5648.78
Half-StepBase Model=Dream-7B-Instruct, Decoding Strategy=Half-Step2026.0423.69235.37
LLaDA + Half-StepBackbone=LLaDA-8B-Instruct2026.0421.67235.37
dKV-CacheBase Model=Dream-7B-Instruct, Decoding Strategy=dKV-Cache2026.0415.211.2956.1
LLaDA + dKV-CacheBackbone=LLaDA-8B-Instruct2026.0414.231.3246.95
Vanilla DreamBase Model=Dream-7B-Instruct, Decoding Strategy=Vanilla2026.0411.82159.15
Vanilla LLaDABackbone=LLaDA-8B-Instruct2026.0410.82148.17