Gemini 2.0 Flash: benchmark results
Google's workhorse Gemini 2.0 model with a 1M context, native tool use and multimodal input/output (December 2024). Provider: Google. Released 2024-12-11. Access: API.
Aggregate Rating 1635 ± 4, rank #342 of 1656 rated models, from 1804 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| APUN-Bench (Real-World) - Coarse Pun Location | 67.52 | Coarse pun-location accuracy (%), the predicted pun word mat | 100 |
| APUN-Bench (Real-World) - Fine Pun Location | 51.03 | Fine pun-location accuracy (%), the predicted pun word exact | 100 |
| APUN-Bench (Real-World) - Heterographic Pun Inference | 62.78 | Heterographic pun-inference accuracy (%), the predicted alte | 100 |
| APUN-Bench (Real-World) - Homophonic Pun Inference | 82.85 | Homophonic pun-inference accuracy (%), the predicted alterna | 100 |
| APUN-Bench (Synthetic) - Homophonic Pun Inference | 92.11 | Homophonic pun-inference accuracy (%), the predicted alterna | 100 |
| APUN-Bench (Synthetic) - Pun Recognition | 80 | Pun-recognition accuracy (%), deciding whether the audio cli | 100 |
| ATAD - Tone and Style Violation | 88 | Accuracy (%) on the T7 tone and style violation task (pick t | 100 |
| AVI-Bench - Audio-Visual Sensation QA | 26.8 | Task score (%) on primitive audio-visual sensation (388 mult | 100 |
| CASU - Entity Extraction | 72.77 | Accuracy (%; four-option multiple-choice questions on semi-s | 100 |
| CASU - Role Inference | 69.05 | Accuracy (%; four-option multiple-choice questions on semi-s | 100 |
| CASU - Scene Description Event Recall | 75 | Event match (%; an LLM judge counts the share of the scene s | 100 |
| CIPHER Cross-Record Inference - CFPB (Chain-of-Thought) | 39.7 | Harmonic answer score (%; per question the harmonic mean of | 100 |
Interactive version: theaggregate.ai/model?slug=gemini-2-0-flash · How It Works · Data refreshed daily, snapshot 2026-10-20.