Which is better, Gemini 3.5 Flash or Gemini 3.6 Flash?

Across 42 shared benchmarks tracked on Vector Wire, Gemini 3.5 Flash leads on 21 and Gemini 3.6 Flash leads on 21 — Gemini 3.5 Flash leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Gemini 3.5 Flash vs Gemini 3.6 Flash

GooglevsGoogle42 shared benchmarks21–21 head-to-head

Benchmark	Gemini 3.5 Flash	Gemini 3.6 Flash
AA Agentic Index	70.4	40.5
AA Intelligence	52	52
AA-LCR	81	79
AA-Omniscience	21.2	22.1
aime_2026	95	96.7
arc_agi_1	92.5	34.5
ARC-AGI-2	72.1	2.6
arena_elo	1477	1481
arena_text_factuality	1466	1470
Artificial Analysis Coding Index	70.1	69.2
charxiv_reasoning_with_tools	84.9	89.4
charxiv_rq	84.2	85.2
coding_arena_elo	1490	1538
critpt	13.1	10.6
DeepSWE	37	49
DeepSWE 1.1	37	49
GDM-MRCR v2 (8-needle) (1M (pointwise))	26.6	54
gdpval	57.8	46.1
GDPval-AA v2	1357	1421
GDPVal-AA v2 (Elo)	1349	1422
GPQA Diamond	92.2	92.8
HLE	42.7	40.8
hmmt_feb_2026	95.5	89.4
livebench_agentic_coding	49	43.4
livebench_coding	78.2	77.9
livebench_data_analysis	64.9	63
livebench_instruction_following	75.6	75.4
livebench_language	84.6	83.9
livebench_math	88.2	86.4
livebench_reasoning	82	85.2
matharena_visual_math_overall	89.9	87.8
MLE-Bench	49.7	63.9
MMMU-Pro	84.3	83.2
mrcr_v2_8needle_128k_average	77.3	91.8
OmniScience Accuracy	51.4	50
OmniScience Non-Hallucination	38.2	44.4
OSWorld-Verified	78.4	83
scicode	53.1	52.7
simpleqa_verified	68.4	68.7
SWE-bench Pro	55.1	58.7
TauBench V3 - Banking	32.2	29.9
Terminal-Bench 2.1	78.7	78

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.