LetTokenBurnBenchHub

Model

Qwen3.5 4B

Alibaba · open weights

qwenlm.github.io

Knowledge & Reasoning

Knowledge & Reasoning

Benchmark	Score	Metric	Harness	Source	Confidence	Link
HouseholdBench	1.3	relmae	N/A	Independent	medium	arxiv.org
LLM4Impact Dataset	1.6	rmse	zero-shot	Independent	medium	arxiv.org
RoleCapBench	0.9	accuracy	N/A	Independent	medium	arxiv.org

Math

Benchmark	Score	Metric	Harness	Source	Confidence	Link
GraphDecide	99.7	accuracy	N/A	Independent	medium	arxiv.org

Coding

Benchmark	Score	Metric	Harness	Source	Confidence	Link
ChartRevise	93.8	exact-edit_rate	N/A	Independent	medium	arxiv.org
Closed-loop evaluation benchmark for embedded coding agents	13.3	pass_rate	N/A	Independent	medium	arxiv.org
Mobility Data Engineering Benchmark	37.3	success_rate	N/A	Independent	medium	arxiv.org

Agentic / Tool Use

Agentic / Tool Use

Benchmark	Score	Metric	Harness	Source	Confidence	Link
SearchDecision-Bench	78.4	accuracy	N/A	Independent	medium	arxiv.org

Multimodal

Benchmark	Score	Metric	Harness	Source	Confidence	Link
DSV-Mem	3.3	accuracy	N/A	Independent	medium	arxiv.org
NextAct	0.3	embedding_score	N/A	Independent	medium	arxiv.org
SpaceCast-Bench	33.7	accuracy	N/A	Independent	medium	arxiv.org
SpaceConflict	55.8	accuracy	N/A	Independent	medium	arxiv.org

Safety & Alignment

Safety & Alignment

Benchmark	Score	Metric	Harness	Source	Confidence	Link
Unified Misuse Monitoring Benchmark	0.9	auc	N/A	Independent	medium	arxiv.org

Multilingual

Benchmark	Score	Metric	Harness	Source	Confidence	Link
HakemBench	0.7	composite	N/A	Independent	medium	arxiv.org

Reproduced by the community