GenMagic logoGenMagic
Sign in

Claude Sonnet 5 vs Gemini 3.7 Flash

Anthropic against Google. Prices, scores and speed below are live.

  • Gemini 3.7 Flash costs 63% less on blended price ($2.25 against $6.00 per 1M tokens, 3 input to 1 output).
  • Gemini 3.7 Flash scores higher on the Artificial Analysis Intelligence Index: 39.4 against 38.4.
  • On the 40 benchmarks both have scores on, Claude Sonnet 5 leads on 4 and Gemini 3.7 Flash on 36.
  • Gemini 3.7 Flash reads more context: 1M tokens against 1M.
  • Gemini 3.7 Flash is newer: listed Aug 13, 2026, against Jun 30, 2026.
Run both on your promptClaude Sonnet 5Gemini 3.7 Flash

Side by side

Claude Sonnet 5Gemini 3.7 Flash
LabAnthropicGoogle
ListedJun 30, 2026Aug 13, 2026
Input per 1M tokens$3.00$1.13
Output per 1M tokens$15.00$5.63
Cached input per 1M$0.30$0.11
Blended (3:1) per 1M$6.00$2.25
Context window1M1M
Max output128K66K
Knowledge cutoff--
Intelligence index38.439.4
Coding index71.576.1
ReasoningYesYes
Reads imagesYesYes
Tool useYesYes
Structured outputYesYes
Measured output speed70 tokens/s-

Every benchmark both are scored on

Each model's best score per benchmark, with its source. Scores from different sources or settings are not always comparable; each links to where it was published.

BenchmarkClaude Sonnet 5Gemini 3.7 Flash
AIME80%Epoch AI97.2%Epoch AI
APEX-Agents54.5%Epoch AI67.8%Epoch AI
Artificial Analysis Agentic Index43.6Artificial Analysis and OpenRouter35.2Artificial Analysis and OpenRouter
Artificial Analysis Coding Index71.5Artificial Analysis and OpenRouter76.1Artificial Analysis and OpenRouter
Artificial Analysis Intelligence Index38.4Artificial Analysis and OpenRouter39.4Artificial Analysis and OpenRouter
Chess Puzzles45%Epoch AI66%Epoch AI
DeepSWE53.8%Epoch AI65.5%Epoch AI
Design Arena: 3D1258 EloDesign Arena1306 EloDesign Arena
Design Arena: Android Native (agentic)1221 EloDesign Arena1239 EloDesign Arena
Design Arena: ASCII Art1204 EloDesign Arena1275 EloDesign Arena
Design Arena: Code1274 EloDesign Arena1306 EloDesign Arena
Design Arena: Data Viz1252 EloDesign Arena1316 EloDesign Arena
Design Arena: Full-stack (agentic)1249 EloDesign Arena1208 EloDesign Arena
Design Arena: Game Dev1275 EloDesign Arena1319 EloDesign Arena
Design Arena: Game Dev (agentic)1172 EloDesign Arena1183 EloDesign Arena
Design Arena: Godot Game Dev (agentic)1225 EloDesign Arena1264 EloDesign Arena
Design Arena: Mobile Apps (agentic)1208 EloDesign Arena1232 EloDesign Arena
Design Arena: SVG1190 EloDesign Arena1283 EloDesign Arena
Design Arena: UI Component1274 EloDesign Arena1294 EloDesign Arena
Design Arena: Web Apps (agentic)1227 EloDesign Arena1208 EloDesign Arena
Design Arena: Website1273 EloDesign Arena1303 EloDesign Arena
DTBench92.5%Epoch AI96.8%Epoch AI
FrontierCode42.7Epoch AI43.6Epoch AI
FrontierMath Tier 4 (v2)29.3%Epoch AI36.6%Epoch AI
FrontierMath Tiers 1-3 (v2)65.6%Epoch AI71.6%Epoch AI
GPQA Diamond85%Artificial Analysis and OpenRouter93.9%Artificial Analysis and OpenRouter
LMArena (Coding)1484 EloLMArena1500 EloLMArena
LMArena (Overall)1443 EloLMArena1489 EloLMArena
LMArena: Creative Writing1416 EloLMArena1487 EloLMArena
LMArena: Hard Prompts1462 EloLMArena1497 EloLMArena
LMArena: Longer Query1463 EloLMArena1494 EloLMArena
LMArena: Math1482 EloLMArena1501 EloLMArena
LMArena: Multi-Turn1453 EloLMArena1485 EloLMArena
LMCA50Epoch AI50.4Epoch AI
Mystery Game Puzzles16%Epoch AI37%Epoch AI
ProofBench77%Epoch AI58%Epoch AI
SciCode54.3%Epoch AI59.8%Epoch AI
SimpleQA Verified32.9%Epoch AI69.2%Epoch AI
tau-bench (Airline)75.3%Artificial Analysis and OpenRouter78.3%Artificial Analysis and OpenRouter
WebDev Arena1537 EloEpoch AI1587 EloEpoch AI

In our tests and guides

Where Claude Sonnet 5 or Gemini 3.7 Flash ran on a real brief, or shows in a live table.

More comparisons

Where these figures come from

Prices are GenMagic's own, what the studio and the API charge, recomputed from each provider's price list every few minutes. The indices are the Artificial Analysis indices, and every benchmark score links to its source, re-read daily. Measured speed is GenMagic's own generation records, in aggregate. See Anthropic models, Google models, every LLM's price and our sources.