Compare to
Discover how Anthropic's Claude Opus 5 and Google's Gemini 3.8 Flash stack up against each other in this comprehensive comparison of two leading AI language models. Released in July 2026 and September 2026 respectively, these models represent significant advancements in artificial intelligence, with Claude Opus 5 offering a 1,000,000-token context window and Gemini 3.8 Flash offering a 1,048,576-token context window.
Explore their capabilities, pricing, and performance metrics, with Claude Opus 5 achieving 1,493 on LMArena Elo and Gemini 3.8 Flash scoring 1,493, making this comparison essential for developers and organizations seeking the right AI solution for their specific needs.
Models Overview
 Claude Opus 5 |  Gemini 3.8 Flash | |
|---|---|---|
Provider The company that provides the model. | Anthropic | |
Context Length Maximum number of tokens the model can process | 1M | 1.05M |
Maximum Output Maximum number of tokens the model can generate in one response | 128K | 65.54K |
Release Date When the model was first released. | 24-07-2026 | 02-09-2026 |
Knowledge Cutoff When the model's training data ends. | 2026-05 | Unknown |
Open Source Whether the model weights are openly available. | FALSE | FALSE |
Pricing Comparison
Compare the pricing of Anthropic's Claude Opus 5 and Google's Gemini 3.8 Flash to determine the most cost-effective solution for your AI needs. Prices are the standard API tier per million tokens, as published by each provider as of September 2026.
| Claude Opus 5 | Gemini 3.8 Flash | |
|---|---|---|
Input Cost Cost per million input tokens | $5 / 1M tokens | $0.75 / 1M tokens |
Output Cost Cost per million tokens generated | $25 / 1M tokens | $3.75 / 1M tokens |
Comparing Benchmarks and Performance
Compare the performances of Anthropic's Claude Opus 5 and Google's Gemini 3.8 Flash on industry benchmarks. Scores are the ones the providers and public leaderboards report; a benchmark neither reports is left out.
| Claude Opus 5 | Gemini 3.8 Flash | |
|---|---|---|
LMArena Elo Crowd-sourced blind preference rating on the LMArena text leaderboard. | 1,493 | 1,493 |
Terminal-Bench 2.1 Agentic tasks completed in a real terminal. | Benchmark not available | 89.4% |
Humanity's Last Exam Expert-written questions at the frontier of human knowledge. | 56.3% | Benchmark not available |
Sources — Claude Opus 5: platform.claude.com, arena.ai, anthropic.com; Gemini 3.8 Flash: ai.google.dev, blog.google, arena.ai, deepmind.google.
GPT-6 Astra
Muse Spark 1.3