Qwen3.6-27B vs Gemma3-27B vs Qwen3.8-27B

New Comparison

Scores are fetched live from the current database on every visit — nothing is cached.

Benchmark Scores Comparison

Qwen3.6-27B Gemma3-27B Qwen3.8-27B

Qwen3.6-27B

27.0B params

Score: 100.0

Gemma3-27B

27.0B params

Score: 100.0

Qwen3.8-27B

27.0B params

Score: 100.0

Coverage matrix

3 models × 90 benchmarks — hover a benchmark name for its category

Full Coverage Matrix
Shading = score relative to the best score among the selected models: Low Mid High Missing (—) The totals row shows how many of the 90 benchmarks each model has scores for.
Benchmark Gemma3-27B Google Qwen3.6-27B Qwen Team (Alibaba Cloud) Qwen3.8-27B Qwen Team (Alibaba Cloud)
HMMT Feb 25 — 93.8 —
DynaMath — 85.6 —
GPQA Diamond 42.4 87.8 89.2
IFBench (prompt loose) — 69.1 79.5
MathVista (mini) — 87.4 —
MMBench EN-DEV v1.1 — 92.3 —
LiveCodeBench v6 29.1 83.9 90.3
SWE-bench Pro — 53.5 61.7
RealWorldQA — 84.1 85.9
Agents' Last Exam (Score) — — 42.9
EmbSpatialBench — 84.6 —
AA-LCR — 73.3 —
MathVision 46 85.1 90
AIME 26 20.8 94.1 —
AndroidWorld — 70.3 81.9
Agents' Last Exam — 10.6 20.4
Toolathlon Verified — — 67.1
RecreationBench — 29.8 47.1
ScreenSpot Pro — 76.1 —
WMDP (Bio) — 84.8 —
MMStar — 81.4 —
DeepSearch QA — 71.1 —
WildClawBench — 43.2 —
NL2Repo — 36.2 42.3
HPCT — 48.7 —
MMMU-Pro 49.7 75.8 —
QwenClawBench — 53.4 —
BabyVision — 28.9 65.7
MMMLU 70.7 — —
Claw-Eval Avg — 72.4 56.9
VideoMMMU — 84.4 —
HMMT Nov 25 — 90.7 —
SWE-MM — 25.7 38.6
MMLU-Pro 67.6 86.2 —
CodeForces 110 — —
SuperGPQA — 66 —
OSWorld-Verified — 63.9 84.3
HLE (with tools) — — 30.8
LVBench — — 72.4
CharXiv (RQ) — 78.4 83.7
MRCR v2 128K (8-needle) 13.5 — —
MMLU-Redux — 93.5 —
NL2Repo-Bench — — 42.3
SkillsBench — 46.6 —
WebArena-Verified — 48.8 64.8
CoWorkBench — 61 70.7
Lab Bench (ProtocolQA) — 69.1 —
SWE-bench Multilingual — 71.3 73.8
MLVU — 86.6 —
Beam128K — 63 —
IFBench — 70.8 79.5
SWE-bench Verified — 77.2 —
IMOAnswerBench — 80.8 —
Humanity's Last Exam — 24 30.8
CountBench — 97.8 —
WMDP (Chem) — 74.8 —
ERQA — 62.5 65.5
Terminal-Bench 2.0 — 59.3 —
QwenSWEBench — 49.3 79
QwenWebBench — 1487 —
TAU2-Bench 16.2 — —
VlmsAreBlind — 97 —
Terminal-Bench 2.1 (Terminus-2) — 63.4 73
OmniDocBench 1.5 0.365 89.4 91.1
OCRBench — 89.4 —
HMMT Feb 26 — 84.3 —
CC-OCR — 81.2 —
RefCOCO (avg) — 92.5 —
JobBench — 21.8 33.4
SciCode (subtask) — 39.8 —
OSWorld 2.0 (Partial) — — 48
OSWorld 2.0 (Binary) — — 19.4
C-Eval — 91.4 —
SimpleVQA — 56.1 —
VideoMME (w sub.) — 87.7 —
SkillsBench Avg5 — 48.2 —
Vision2Web — 45 62.9
TauBench V3 Banking — 16.7 —
Claw-Eval Pass^3 — 60.6 57.4
MCP-Atlas — 62.5 —
V-Star — 94.7 —
DeepSWE 1.1 — 13.3 42.2
MBCT — 45.9 —
BigBench Extra Hard 19.3 — —
VCT — 33.7 —
Gaia2 — 40 —
RefSpatialBench — 70 —
MMMU — 82.9 —
GDPVal-AA v2 — 1141 —
MVBench — 75.5 —
Covered 12/90 78/90 35/90

Fully comparable benchmarks

Scored by all 3 selected models (4)

Benchmark Category Qwen3.6-27B Gemma3-27B Qwen3.8-27B
GPQA Diamond stem_reasoning 87.8 42.4 89.2
LiveCodeBench v6 stem_reasoning 83.9 29.1 90.3
MathVision vision_language 85.1 46 90
OmniDocBench 1.5 document_understanding 89.4 0.365 91.1

Partial coverage benchmarks

Scored by at least one but not all selected models (86) — missing scores show as —

Benchmark Category Qwen3.6-27B Gemma3-27B Qwen3.8-27B
HMMT Feb 25 1/3 models stem_reasoning 93.8 — —
DynaMath 1/3 models vision_language 85.6 — —
IFBench (prompt loose) 2/3 models instruction_following 69.1 — 79.5
MathVista (mini) 1/3 models vision_language 87.4 — —
MMBench EN-DEV v1.1 1/3 models vision_language 92.3 — —
SWE-bench Pro 2/3 models coding_agent 53.5 — 61.7
RealWorldQA 2/3 models vision_language 84.1 — 85.9
Agents' Last Exam (Score) 1/3 models general_agent — — 42.9
EmbSpatialBench 1/3 models spatial_intelligence 84.6 — —
AA-LCR 1/3 models long_context 73.3 — —
AIME 26 2/3 models stem_reasoning 94.1 20.8 —
AndroidWorld 2/3 models general_agent 70.3 — 81.9
Agents' Last Exam 2/3 models general_agent 10.6 — 20.4
Toolathlon Verified 1/3 models general_agent — — 67.1
RecreationBench 2/3 models general_agent 29.8 — 47.1
ScreenSpot Pro 1/3 models general_agent 76.1 — —
WMDP (Bio) 1/3 models safety 84.8 — —
MMStar 1/3 models vision_language 81.4 — —
DeepSearch QA 1/3 models general_agent 71.1 — —
WildClawBench 1/3 models coding_agent 43.2 — —
NL2Repo 2/3 models coding_agent 36.2 — 42.3
HPCT 1/3 models safety 48.7 — —
MMMU-Pro 2/3 models vision_language 75.8 49.7 —
QwenClawBench 1/3 models coding_agent 53.4 — —
BabyVision 2/3 models vision_language 28.9 — 65.7
MMMLU 1/3 models multilingual — 70.7 —
Claw-Eval Avg 2/3 models coding_agent 72.4 — 56.9
VideoMMMU 1/3 models video_understanding 84.4 — —
HMMT Nov 25 1/3 models stem_reasoning 90.7 — —
SWE-MM 2/3 models coding_agent 25.7 — 38.6
MMLU-Pro 2/3 models knowledge 86.2 67.6 —
CodeForces 1/3 models stem_reasoning — 110 —
SuperGPQA 1/3 models knowledge 66 — —
OSWorld-Verified 2/3 models general_agent 63.9 — 84.3
HLE (with tools) 1/3 models stem_reasoning — — 30.8
LVBench 1/3 models video_understanding — — 72.4
CharXiv (RQ) 2/3 models document_understanding 78.4 — 83.7
MRCR v2 128K (8-needle) 1/3 models long_context — 13.5 —
MMLU-Redux 1/3 models knowledge 93.5 — —
NL2Repo-Bench 1/3 models coding_agent — — 42.3
SkillsBench 1/3 models general_agent 46.6 — —
WebArena-Verified 2/3 models general_agent 48.8 — 64.8
CoWorkBench 2/3 models general_agent 61 — 70.7
Lab Bench (ProtocolQA) 1/3 models safety 69.1 — —
SWE-bench Multilingual 2/3 models coding_agent 71.3 — 73.8
MLVU 1/3 models video_understanding 86.6 — —
Beam128K 1/3 models long_context 63 — —
IFBench 2/3 models instruction_following 70.8 — 79.5
SWE-bench Verified 1/3 models coding_agent 77.2 — —
IMOAnswerBench 1/3 models stem_reasoning 80.8 — —
Humanity's Last Exam 2/3 models stem_reasoning 24 — 30.8
CountBench 1/3 models spatial_intelligence 97.8 — —
WMDP (Chem) 1/3 models safety 74.8 — —
ERQA 2/3 models spatial_intelligence 62.5 — 65.5
Terminal-Bench 2.0 1/3 models coding_agent 59.3 — —
QwenSWEBench 2/3 models coding_agent 49.3 — 79
QwenWebBench 1/3 models coding_agent 1487 — —
TAU2-Bench 1/3 models general_agent — 16.2 —
VlmsAreBlind 1/3 models vision_language 97 — —
Terminal-Bench 2.1 (Terminus-2) 2/3 models coding_agent 63.4 — 73
OCRBench 1/3 models document_understanding 89.4 — —
HMMT Feb 26 1/3 models stem_reasoning 84.3 — —
CC-OCR 1/3 models document_understanding 81.2 — —
RefCOCO (avg) 1/3 models spatial_intelligence 92.5 — —
JobBench 2/3 models general_agent 21.8 — 33.4
SciCode (subtask) 1/3 models stem_reasoning 39.8 — —
OSWorld 2.0 (Partial) 1/3 models general_agent — — 48
OSWorld 2.0 (Binary) 1/3 models general_agent — — 19.4
C-Eval 1/3 models knowledge 91.4 — —
SimpleVQA 1/3 models vision_language 56.1 — —
VideoMME (w sub.) 1/3 models video_understanding 87.7 — —
SkillsBench Avg5 1/3 models coding_agent 48.2 — —
Vision2Web 2/3 models vision_language 45 — 62.9
TauBench V3 Banking 1/3 models general_agent 16.7 — —
Claw-Eval Pass^3 2/3 models coding_agent 60.6 — 57.4
MCP-Atlas 1/3 models general_agent 62.5 — —
V-Star 1/3 models vision_language 94.7 — —
DeepSWE 1.1 2/3 models coding_agent 13.3 — 42.2
MBCT 1/3 models safety 45.9 — —
BigBench Extra Hard 1/3 models stem_reasoning — 19.3 —
VCT 1/3 models safety 33.7 — —
Gaia2 1/3 models general_agent 40 — —
RefSpatialBench 1/3 models spatial_intelligence 70 — —
MMMU 1/3 models vision_language 82.9 — —
GDPVal-AA v2 1/3 models general_agent 1141 — —
MVBench 1/3 models video_understanding 75.5 — —