Gemini 3.8 Flash

4 runs · 4 datasets · 1 model

slug: gemini-3-8-flash

0.701

Best SPS · opinionsqa

Disaggregated subgroup scorecard. Each card below is one published run for this vendor; expand the question-type and demographic-subgroup sections to see the matrix beneath the headline SPS. Where coverage permits, 95% CI bands accompany the point estimate.

What each cell measures

Each cell compares this vendor's synthetic responses against real human respondents in that demographic subgroup — published survey ground truth, not a model's guess about the subgroup. Higher = closer to how that real subgroup actually answered.

Demographic conditioning here is not stereotyping: every conditioned score is checked against what real subgroup members said, not against assumptions about them — and where the vendor's output diverges from the real subgroup, the score drops.

globalopinionqa — ground truth: Durmus et al. 2023, Anthropic — llm_global_opinions
gss — ground truth: NORC at the University of Chicago — General Social Survey
opinionsqa — ground truth: Santurkar et al., ICML 2023 — Whose Opinions Do LLMs Reflect? (derived from Pew American Trends Panel)
subpop — ground truth: Suh et al., ACL 2025 — SubPOP: Subpopulation-Level Opinion Prediction

Low-n cells: cells with n < 30 are shown muted and tagged low n — suggestive only instead of color-graded; they are reported for transparency, not as findings.

CIs: "no CI — single run" marks point estimates from a single run with no confidence interval yet; treat the uncertainty as unknown, never zero.

Columns: Score = distributional parity (p_dist) restricted to the subgroup · p_cond = conditioning strength vs the unconditioned baseline · N = questions answered under that conditioning · Cov. = coverage dot derived from N (green = high, ≥100 · amber = medium, 50–99 · red = low, <50). Topic tables: SPS = Survey Parity Score for the topic · p_dist = 1 − mean(JSD) · p_rank = (1 + mean(τ)) / 2 · p_refuse = 1 − mean(|R_model − R_human|).

Multiple comparisons: with this many subgroup cells, a few extreme cells are expected by chance alone — read patterns across a dimension, not single cells.

globalopinionqa raw

raw--gemini-3.8-flash--tdefault--tplcurrent--8f94fc85

0.694 ± 0.040

SPS · 95% CI [0.653, 0.733] · n = 100

Question-type breakdown (6 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Health & Science low n — suggestive only	0.887	0.775	1.000	1.000	1
General Attitudes low n — suggestive only	0.740	0.741	0.739	0.966	4
Trust & Wellbeing low n — suggestive only	0.692	0.580	0.805	0.983	2
International Relations & Security	0.602	0.594	0.610	0.926	60
Economy & Work low n — suggestive only	0.549	0.599	0.500	0.949	3
Politics & Governance	0.487	0.464	0.510	0.926	30

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on globalopinionqa. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

gss raw

raw--gemini-3.8-flash--tdefault--tplcurrent--b56a1300

0.531 ± 0.036

SPS · 95% CI [0.497, 0.569] · n = 75

Question-type breakdown (7 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Social Values & Religion	0.422	0.386	0.458	0.977	14
General Attitudes	0.354	0.348	0.359	0.960	10
Politics & Governance low n — suggestive only	0.328	0.314	0.341	0.975	9
International Relations & Security	0.288	0.270	0.305	0.974	33
Health & Science low n — suggestive only	0.243	0.273	0.213	0.967	2
Economy & Work low n — suggestive only	0.140	0.062	0.217	0.981	4
Trust & Wellbeing low n — suggestive only	0.122	0.116	0.128	0.951	3

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on gss. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

opinionsqa raw

raw--gemini-3.8-flash--tdefault--tplcurrent--cb4842bd

0.701 ± 0.037

SPS · 95% CI [0.665, 0.739] · n = 100

Question-type breakdown (9 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Media & Information low n — suggestive only	0.851	0.841	0.862	0.996	2
Technology & Digital Life low n — suggestive only	0.774	0.772	0.777	0.995	4
Economy & Work	0.699	0.693	0.705	0.995	21
General Attitudes	0.531	0.517	0.546	0.994	35
Health & Science low n — suggestive only	0.521	0.451	0.591	0.997	1
Social Values & Religion	0.481	0.439	0.522	0.993	15
International Relations & Security	0.459	0.456	0.463	0.994	20
Politics & Governance low n — suggestive only	0.386	0.271	0.500	0.996	1
Trust & Wellbeing low n — suggestive only	0.106	0.028	0.184	0.994	1

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on opinionsqa. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

subpop raw

raw--gemini-3.8-flash--tdefault--tplcurrent--37e38f3f

0.600 ± 0.040

SPS · 95% CI [0.559, 0.639] · n = 100

Question-type breakdown (9 topics)

Topic	SPS	p_dist	p_rank	p_refuse	N
Technology & Digital Life	0.654	0.624	0.684	0.993	16
Economy & Work	0.629	0.612	0.647	0.992	13
General Attitudes	0.465	0.444	0.486	0.936	17
International Relations & Security	0.439	0.399	0.479	0.991	15
Health & Science low n — suggestive only	0.320	0.346	0.295	0.994	6
Social Values & Religion	0.194	0.150	0.239	0.985	27
Media & Information low n — suggestive only	0.145	0.078	0.211	0.978	1
Politics & Governance low n — suggestive only	0.118	0.069	0.168	0.981	4
Identity & Demographics low n — suggestive only	0.085	0.023	0.146	0.995	1

Topics with N < 10 are muted and tagged "low n — suggestive only": too few questions for a stable 3-decimal score.

Not yet measured

This vendor has no demographic-conditioned runs for age, geography, education, or any other subgroup dimension on subpop. No cells are fabricated — scores appear here only when a conditioned run actually measured them.

How to submit a demographic-conditioned run →

No demographic conditioning data has been published for this vendor yet. The question-type matrix above shows topic-level parity; subgroup rows fill in once Althing-style conditioned runs land.

← Back to leaderboard