Updated

A large pleated fan of terracotta clay with ivory paper inlays on a dark background, abstract hero artwork for Claude Opus 5

Anthropic

Claude Opus 5

Anthropic's Opus 5, close to Fable 5 at half the price, with thinking you can turn off.

Chat with Claude Opus 5 in Super AgentView pricing

What is Claude Opus 5?

Claude Opus 5 is the Opus model Anthropic released on July 24, 2026, and the predecessor of Claude Opus 5.5.

Anthropic said it comes close to the intelligence of Claude Fable 5 at half the price and more than doubles Opus 4.8's score on Frontier-Bench v0.1. It is much better at checking its own work and iterating until a task succeeds, including root-cause debugging. It also beat Opus 4.8 on every life sciences evaluation Anthropic ran and produces stronger visual outputs. Anthropic now recommends Opus 5.5 for new work, but Opus 5 can still run with thinking off.

Provider
Anthropic
Context window
1M tokens
Input
Text, Image
Reasoning
Off to max
Max output
128K tokens
Knowledge cutoff
May 2026
Released
Jul 24, 2026

How Claude Opus 5 performs

Published by Anthropic

Results Anthropic reported for Claude Opus 5. Each lab runs its own setup, so compare these with other models only inside the same table.

Benchmark results Anthropic published for Claude Opus 5
BenchmarkScoreSetting
Frontier-Bench v0.143.3%
ARC-AGI-330.2%
BrowseComp90.8%
Humanity's Last Exam56.3%no tools
Humanity's Last Exam64.7%with tools
OSWorld 2.070.6%
DeepSWE v1.168.8%
FrontierCode v1.1 (Main)53.4%
AutomationBench26.0%
HealthBench Professional59.8%

Source: anthropic.com

What you can do with Claude Opus 5

  • Thinking can be turned off, unlike Opus 5.5

  • More than double Opus 4.8's score on Frontier-Bench v0.1

  • Checks its own work and keeps iterating until a task succeeds

  • Better than Opus 4.8 on every life sciences test Anthropic ran

How to prompt Claude Opus 5

  • Ask it to verify the result before reporting back; Anthropic says it is much stronger at checking its own work.
  • Turn thinking off for quick edits and back on for planning.
  • Describe the final deliverable in detail so it knows when the task is done.

Where Claude Opus 5 fits best

  • Complex builds and long plans

  • Debugging that needs a root cause

  • Science and research questions

  • Reviewing work for mistakes

How Claude Opus 5 compares

Claude Opus 5 compared with other chat models on Flashloop
ModelLMArena textLMArena codingContext windowReasoning
Claude Opus 51491#131528#231M tokensOff to max
Gemini 3.7 Flash1490#161520#361M tokensLow, medium, high
Kimi K31487#181539#111M tokensOff, low, medium, high, max
Gemini 3.1 Pro Preview1487#191522#321M tokensLow, medium, high
GPT 5.6 Sol1485#211534#171.05M tokensOff to max

Good to know before you start

  • Anthropic lists it as legacy and recommends Opus 5.5
  • Anthropic's API allows thinking off only at high effort or below, not at extra high or max
  • No media output: Super Agent turns its prompts into images and video with Flashloop's generation models

Start chatting with Claude Opus 5

Chat with Claude Opus 5 in Super Agent

  1. Select Claude Opus 5

    Open Super Agent on Flashloop and choose Claude Opus 5 in the model picker.

  2. Enter your prompt

    Ask for what you need: an answer, a plan, or the images and videos you want the agent to make.

  3. Send and keep going

    Send the message. Claude Opus 5 answers in Super Agent and runs any generations you asked for.

Questions about Claude Opus 5

Is Claude Opus 5 still worth using after Opus 5.5?

Opus 5.5 is newer, and Anthropic recommends it for most work. Opus 5 is still useful if you want an Opus model with thinking off, which Opus 5.5 does not allow.

What is Claude Opus 5 good at?

Anthropic points to self-verification and iteration, root-cause debugging, life sciences and visual output. It also set a new high on Frontier-Bench v0.1 at launch.

What reasoning levels does Claude Opus 5 support?

Off, low, medium, high, extra high and max. Super Agent starts it at medium.

What is Claude Opus 5's context window?

It holds 1M tokens and accepts text and images.

More chat models on Flashloop

Browse all models