AI model matchup

Gemini 3.6 Flash vs Claude Sonnet 4.6

A side-by-side look at published capability scores, listed API pricing, speed, context, and license - to help you decide which trade-offs fit your work.

OpenAI

GPT-5.6 Sol

Rank #1 • Score 57.2
Comparison Metrics
Speed: 89c/s vs 68c/s
Pricing: $7.78 vs $7.22
Anthropic

Claude Opus 5

Rank #2 • Score 56.2

Gemini 3.6 FlashvsClaude Sonnet 4.6

Claude Sonnet 4.6 leads the RankLLMs Index by 1.3 points (49.8 vs 48.5). Verified API Pricing: $0.75 (Gemini 3.6 Flash) vs $3.00 (Claude Sonnet 4.6) per 1M tokens.

Capability profile

Gemini 3.6 FlashClaude Sonnet 4.6

Speed duel

tps
Gemini 3.6 Flash165
Claude Sonnet 4.682

Price duel

$/1M
Gemini 3.6 Flash$0.75
Claude Sonnet 4.6$3.00

Shorter bar = cheaper. Blended per 1M tokens.

MetricGemini 3.6 FlashClaude Sonnet 4.6
RankLLMs Overall Score48.5 pts 49.8 pts ▲
Reasoning52.0 ▲48.8
Coding66.8 ▲64.5
Agentic Tool Use38.5 78.5 ▲
Code Arena Elo1538 ▲1395
Inference Speed165 tps ▲82 tps
API Price (per 1M tokens)$0.75 ▲$3.00
Context Window1M 1M
LicenseProprietary Proprietary
Global Rank#40 #37 ▲

Decision notes

What the numbers suggest

Claude Sonnet 4.6 leads the RankLLMs overall score by 1.3 points. Gemini 3.6 Flash has the lower listed API rate ($0.75 per 1M tokens).

Treat the overall index as a starting point, then prioritize the benchmark that resembles your task. Published scores and provider pricing can change; this comparison does not represent a first-party test of either model.

Overall index
Claude Sonnet 4.6 leads
48.5 to 49.8 pts
Coding score
Gemini 3.6 Flash leads
66.8 to 64.5 pts
Reasoning score
Gemini 3.6 Flash leads
52.0 to 48.8 pts
Agentic score
Claude Sonnet 4.6 leads
38.5 to 78.5 pts

Choose by workload

Which profile fits your use?

Google

Consider Gemini 3.6 Flash when

your workload gives more weight to the higher listed coding score, the higher listed reasoning score, the lower listed API rate. Compare those signals with your own task mix, latency needs, and budget.

View Gemini 3.6 Flash profile

Anthropic

Consider Claude Sonnet 4.6 when

your workload gives more weight to the higher overall index score. Compare those signals with your own task mix, latency needs, and budget.

View Claude Sonnet 4.6 profile

Compare a different pairRead editorial comparisons