LetTokenBurnBenchHub

Model

Qwen3.8 Flash

Alibaba · open weights

qwenlm.github.io

Knowledge & Reasoning

Knowledge & Reasoning

Benchmark	Score	Metric	Harness	Source	Confidence	Link
ChronoCulture	68.3	both_periods_accurate_(%)	N/A	Independent	medium	arxiv.org
InsClaimBench	76.6	accuracy	N/A	Independent	medium	arxiv.org

Coding

Benchmark	Score	Metric	Harness	Source	Confidence	Link
MetaKernelBench	35.1	paired_lift	cute dsl	Independent	medium	arxiv.org

Agentic / Tool Use

Agentic / Tool Use

Benchmark	Score	Metric	Harness	Source	Confidence	Link
BOTTLED	0.3	f1	N/A	Independent	medium	arxiv.org
BrickBench	1048.0	elo	N/A	Independent	medium	arxiv.org
CineRecon	42.7	score	qwen code 0.24.1	Independent	medium	arxiv.org
EngiWorld	16.1	engiscore	N/A	Independent	medium	arxiv.org
FDE-Gym	0.1	score	N/A	Independent	medium	arxiv.org
GraceBench	0.6	recovery_ability	N/A	Independent	medium	arxiv.org
PhysCodeGym	0.3	s_tot	claude code	Independent	medium	arxiv.org
ToolBottleneckBench	30.6	success_rate	toolbottleneckbench	Independent	medium	arxiv.org
WorldAuditBench	12.7	success_rate	vla–vlm auditors	Independent	medium	arxiv.org
WorldAuditBench	8.5	success_rate	vlm auditors	Independent	medium	arxiv.org

Multimodal

Benchmark	Score	Metric	Harness	Source	Confidence	Link
EMBER-Bench	42.4	accuracy	N/A	Independent	medium	arxiv.org
GDP.pdf	16.6	gdp.pdf_score	N/A	Independent	medium	surgehq.ai
HSS	30.9	pass@1	N/A	Independent	medium	arxiv.org
MVGraph	50.4	accuracy	N/A	Independent	medium	arxiv.org
OmniCompass	43.6	cs_avg	N/A	Independent	medium	arxiv.org
PlotGround-1k	42.6	accuracy	N/A	Independent	medium	arxiv.org

Reproduced by the community