Atlas

Models

← All models

Gemini 2.0 Flash-Lite

Google · 2025-02-25

Gemini 2.0 Flash-Lite is Google's Gemini 2.0 model optimized for speed, scale, and cost efficiency; the gemini-2.0-flash-lite API model became generally available on February 25, 2025. It uses the series' sparse Mixture-of-Experts Transformer architecture, accepts text, image, audio, and video inputs with a 1,048,576-token context window, and returns up to 8,192 text tokens. The alias pointed to the stable gemini-2.0-flash-lite-001 version, which has no distinct thinking mode; Google shut down both identifiers on June 1, 2026.

Benchmark scores

BenchmarkScore
A-Fantasia65.0
A-Fantasia (Valid Attempts with Retries)65.0
A-Fantasia Backwards Spelling97.0
A-Fantasia Chess22.0
A-Fantasia Chess (Valid Attempts with Retries)22.0
A-Fantasia Cube (Valid Attempts with Retries)76.0
A-Fantasia Cube Rotation76.0
A-Fantasia Spelling (Valid Attempts with Retries)97.0
Artificial Analysis Progress Index v1168
Capability121.7
ChessBench Accuracy-3.7
ChessBench Coherence14.2
ChessBench Elo171
DTBench52.5
ForecastBench Baseline Dataset Brier Index56.6
ForecastBench Baseline Market Brier Index56.4
ForecastBench Baseline Overall Brier Index56.1
ForecastBench Preliminary Dataset Brier Index56.5
ForecastBench Tournament Dataset Brier Index56.6
ForecastBench Tournament Market Brier Index64.9
ForecastBench Tournament Overall Brier Index60.5
GPQA Diamond (Artificial Analysis avg@5)53.5
HardGeoBench <=100mi38.8
HardGeoBench <=10mi22.5
HardGeoBench Avg Distance1216
HardGeoBench Country Accuracy55.0
Humanity's Last Exam (Preview)4.6
Humanity's Last Exam (Text-Only)3.4
Humanity's Last Exam Text Only (Preview)4.4
LisanBench Difficulty-Weighted38.8
Loading Atlas data…