Gemini 3.6 Flash holds steady overall but drops 14% on chart understanding
llama_index · x · 2026-07-22
A ParseBench benchmark of Gemini 3.6 Flash and Gemini 3.5 Flash Lite suggests Google’s newer Flash variants have not clearly improved document understanding.
Key findings
- Gemini 3.6 Flash is roughly on par overall, but drops 14% on chart understanding versus Gemini 3.5 Flash.
- Gemini 3.5 Flash Lite improves layout detection by 11%, but regresses on tables by about 12%.
- The author argues the Flash line appears to have been post-trained more for coding and reasoning, creating a plateau in visual/document recognition.
The benchmark image shows the same pattern: stronger reasoning-oriented models can outperform on agentic or layout tasks, while visual understanding lags in some subsets.
Related event: Gemini 3.6 Flash Review: Faster and Cheaper, But Not Smarter(16 posts)→
More from Models
- BullshitBench update: GPT-6-Astra beats all prior OpenAI models but still trails Anthropic — scaling01 · 2026-09-11
- Astra Scores 83% on GauntletBench, First Computer-Use Agent to Beat Human Baseline — ducha_aiki · 2026-09-11
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11
- Looking for a classifier of software engineering task shapes to pick models per task — StewartalsopIII · 2026-09-11
- DeepSeek V4 Pro API to continue after Sept 2026, billing unchanged — teortaxesTex · 2026-09-11
- DeepSeek V4.1 Flash Hits 98% of GPT-6 Astra's Score at 1.4% of the Cost in Third-Party Benchmark — ayushtweetshere · 2026-09-11