Google: Gemini 3.5 Flash Lite

Google

budget

Capabilities

VisionAudioToolsStructured OutputExtended ThinkingLong ContextCode

Strengths

Exceptional value: budget pricing for strong multimodal performance
1M token context window enables long-document analysis
Robust vision and audio capabilities at low cost
Excellent instruction-following and creative writing

Weaknesses

Slow inference speed (409 tok/s) limits real-time applications
Moderate agentic and reasoning capabilities
Web development performance noticeably weaker than core tasks
Proprietary, closed-source weights restrict customisation

Pricing

Input / 1M tokens

$0.3

Output / 1M tokens

$2.5

Context window

1,048.576k

Transparency

Open weights

0.0 / 10

Open training data

0.0 / 10

Open methodology

3.5 / 10

Licence openness

1.5 / 10

Provider disclosure

5.5 / 10

FMTI company score

N/A

Composite:4.5 / 10

Closed-source model with limited public methodology disclosure. Google publishes some research on Gemini architecture but training data composition remains proprietary. Standard commercial licensing applies.

Sustainability

Inference energy

N/A

Training footprint

N/A

Provider infrastructure

7.5 / 10

Composite:6.2 / 10

Google operates strong renewable energy commitments across data centres. Lite variant optimised for efficiency reduces per-inference energy footprint. Training footprint unknown but likely substantial for flagship family.

Task Fitness

question answering

61%

code

62%

mathematics

85%

business communication

84%

vision

75%

analysis

68%

extraction

87%

long-form generation

85%

research

81%

reasoning

53%

summarisation

85%

translation

80%

conversation

86%