DataLearner logo

Gemini 3.5 Flash-LitevsDeepSeek-V4-Flash

Across 3 shared benchmarks, Gemini 3.5 Flash-Lite leads overall: Gemini 3.5 Flash-Lite wins 2, DeepSeek-V4-Flash wins 1, with 0 ties and an average score difference of -7.77.

Google Deep Mind
Gemini 3.5 Flash-Lite

Google Deep Mind · 2026-07-21 · Multimodal model

DeepSeek-AI
DeepSeek-V4-Flash

DeepSeek-AI · 2026-04-24 · Reasoning model

Gemini 3.5 Flash-Lite2 wins(67%)(33%)1 winDeepSeek-V4-Flash

Benchmark scores

Grouped by capability, sorted by largest gap within each. 3 shared benchmarks.

AI Agent - Tool Usage

DeepSeek-V4-Flash 1/1
BenchmarkGemini 3.5 Flash-LiteDeepSeek-V4-FlashDiff
Terminal-Bench 2.15443 / 45Thinking (With Tools)82.7017 / 45Max (With Tools)-28.70

Coding and Software Engineer

Gemini 3.5 Flash-Lite 1/1
BenchmarkGemini 3.5 Flash-LiteDeepSeek-V4-FlashDiff
SWE-Bench Pro - Public54.2034 / 57Thinking (With Tools)49.1047 / 57Normal (With Tools)+5.10

Writing and Creative Capabilities

Gemini 3.5 Flash-Lite 1/1
BenchmarkGemini 3.5 Flash-LiteDeepSeek-V4-FlashDiff
Creative Writing1,55641 / 99Normal (No Tools)1,55642 / 99Normal (No Tools)+0.30

Specs

FieldGemini 3.5 Flash-LiteDeepSeek-V4-Flash
PublisherGoogle Deep MindDeepSeek-AI
Release date2026-07-212026-04-24
Model typeMultimodal modelReasoning model
ArchitectureDenseMoE
ParametersNot available284B
Context length1M1M
Max output64K384K

API pricing

Prices use DataLearner records when available; missing fields are not inferred.

ItemGemini 3.5 Flash-LiteDeepSeek-V4-Flash
Text input$0.3 / 1M tokens$0.14 / 1M tokens
Text output$2.5 / 1M tokens$0.28 / 1M tokens
Cache read$0.03 / 1M tokens$0.0028 / 1M tokens

Summary

  • Gemini 3.5 Flash-Liteleads in:Coding and Software Engineer (1/1), Writing and Creative Capabilities (1/1)
  • DeepSeek-V4-Flashleads in:AI Agent - Tool Usage (1/1)

On average across the 3 shared benchmarks, DeepSeek-V4-Flash scores 7.77 higher.

Largest single-benchmark gap: Terminal-Bench 2.1 — Gemini 3.5 Flash-Lite 54 vs DeepSeek-V4-Flash 82.70 (-28.70).

Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.