benchgap
Other · model

LLaDA2.2-flash benchmark scores

As of 2026-11-22, LLaDA2.2-flash (Other) has measured scores on 8 benchmarks and estimated scores on 40 more.

BenchmarkScoreSource
AA-SciCode34.8%estimated ± 3.6 pp, high confidence
GDPval-AA6.9%estimated ± 5.5 pp, medium confidence
LiveCodeBench v677.4%estimated ± 2.7 pp, high confidence
SciCode32.3%estimated ± 3.6 pp, high confidence
BrowseComp61.3%estimated ± 7.8 pp, medium confidence
HLE w/ tools32.4%estimated ± 5.5 pp, low confidence
VITA-Bench32.6%estimated ± 9.9 pp, low confidence
τ²-bench results80.3%measured
AA Coding Index24.4%estimated ± 5.3 pp, medium confidence
APEX-Agents3.1%estimated ± 12.6 pp, low confidence
APEX-Agents-AA17.0%estimated ± 5.5 pp, medium confidence
DeepSearchQA81.1%estimated ± 7.3 pp, low confidence
SWE-bench Verified49.3%measured
BFCL v460.8%measured
CyberGym56.7%estimated ± 9.8 pp, medium confidence
JobBench23.6%estimated ± 7.9 pp, medium confidence
SWE-bench Pro30.1%measured
WideResearch70.9%estimated ± 6.7 pp, low confidence
AA Agentic Index19.5%estimated ± 12.7 pp, low confidence
Vals LiveCodeBench20.8%estimated ± 4.7 pp, low confidence
AA Tau3 Banking26.0%estimated ± 4.5 pp, low confidence
CursorBench 3.23.9%estimated ± 3.6 pp, medium confidence
FrontierCode 1.1 Main22.1%estimated ± 5.0 pp, medium confidence
FrontierSWE v20.0%estimated ± 8.3 pp, low confidence
OSWorld-Verified71.1%estimated ± 3.8 pp, medium confidence
Vals SWE-bench34.8%estimated ± 6.6 pp, low confidence
AA ITBench29.2%estimated ± 4.4 pp, medium confidence
DeepSWE30.3%estimated ± 11.2 pp, low confidence
DRACO56.4%estimated ± 4.0 pp, low confidence
FrontierCode 2.017.9%estimated ± 6.7 pp, low confidence
PostTrainBench v1.110.6%estimated ± 4.4 pp, medium confidence
SWE Multilingual25.0%measured
Toolathlon-Verified36.9%estimated ± 8.8 pp, low confidence
VulcanBench Frontier v456.2%estimated ± 8.3 pp, low confidence
Vibe Code Bench0.0%estimated ± 8.4 pp, low confidence
Claw-Eval64.2%measured
DeepPlanning32.3%estimated ± 13.1 pp, low confidence
MCP Atlas46.2%measured
MCP-Tasks73.8%estimated ± 0.6 pp, medium confidence
NL2Repo10.5%estimated ± 6.1 pp, low confidence
QwenClawBench53.7%estimated ± 2.6 pp, high confidence
Toolathlon37.2%estimated ± 7.5 pp, medium confidence
LiveCodeBench Pro54.0%estimated ± 6.0 pp, low confidence
React Native Evals65.8%estimated ± 2.7 pp, low confidence
ResearchClawBench17.5%estimated ± 1.3 pp, high confidence
PostTrainBench9.8%estimated ± 3.7 pp, medium confidence
LiveCodeBench57.6%estimated ± 14.6 pp, low confidence
PinchBench81.7%measured