Is Claude Opus 5 Better Than Claude Fable 5?

TL;DR

Claude Opus 5 is the creator’s preferred model for difficult coding tasks because it generally outperformed Fable 5 in his tests while costing less overall. Its main drawbacks are an overly verbose personality, restrictive Claude usage limits, and a weaker coding harness than the ChatGPT app, so the creator still chooses Fable 5 for planning and ChatGPT 5.6 for daily use.

Transcript

Claude Opus 5 has dropped. It has happened and it has totally blown my mind. It is better than Claude Fable 5. It is a fraction of the price. It is significantly faster, but it has three insane weaknesses I'm about to go over that change absolutely everything. But before we get into those three deal breakers with Claude Opus 5, let's talk about the... Read More

Key Insights

• Claude Opus 5 is the creator’s winner across five custom tests because it won or effectively prevailed in four comparisons, while Fable 5 won the bridge simulation by holding slightly more weight. These results reflect the creator’s own benchmark rather than an independent standardized evaluation.
• The roller coaster test favored Opus 5 because its simulation included a detailed track, trees, buildings, sky, clouds, and a working first-person ride. The creator reports that it required about 32,000 tokens and $0.82 in credits, while Fable 5 produced a less detailed result at a higher cost.
• The website recreation test favored Opus 5 because its hand-built approximation of Apple’s page represented devices and page components more convincingly. Neither recreation was described as perfect, but the creator judged Fable 5’s phones, computers, tablet, watch, people, and cards as substantially less accurate.
• The agentic document test showed that Opus 5 found five of eight scavenger-hunt items across documents, PDFs, and spreadsheets. Fable 5 did not finish because an Anthropic content restriction mistakenly treated the task as related to cybersecurity, so the creator excluded its incomplete attempt from direct scoring.
• The debugging test favored Opus 5 on cost because it addressed bugs collected from about 15 open-source GitHub repositories for roughly one dollar less than Fable 5. Opus took longer and used about the same number of tokens, illustrating that its advantage in this test was economic rather than speed.
• The bridge simulation was the only test won by Fable 5 because its design supported slightly more simulated weight. However, the creator says Fable 5 cost approximately twice as much for that test, making its relatively small performance advantage less attractive from a cost-efficiency perspective.
• Opus 5’s conversational style is the creator’s most immediate usability complaint because it produces long, complicated, and sometimes unfocused responses. He says this behavior can affect coding tasks by prompting the model to perform many additional actions before addressing the requested bug, so he recommends explicitly requesting simple and concise communication.
• The creator’s preferred model stack assigns Opus 5 to difficult coding, Fable 5 to major planning and brainstorming, and ChatGPT 5.6 to everyday work. His choices balance perceived intelligence against conversational focus, usage limits, interface quality, and voice functionality instead of treating benchmark performance as the only criterion.
• More videos from this channel:
  • How to Use ChatGPT Voice to Orchestrate Your Work
  • Is Grok Bot Better Than Hermes and OpenClaw?
  • Is Grok 4.6 the Best Low-Cost Coding Model?
• More from this channel: Alex Finn

Questions & Answers

Q: Is Claude Opus 5 better than Claude Fable 5?

Claude Opus 5 is better for most of the creator’s tested coding and generation tasks. It produced stronger results in the roller coaster simulation, website recreation, document scavenger hunt, and debugging comparison, while costing $6 overall versus $7.50 for Fable 5. Fable 5 built the slightly stronger simulated bridge and remained preferable to the creator for planning and conversation.

Q: What benchmarks were used to compare Opus 5 and Fable 5?

The creator used five custom tests: building a three-dimensional roller coaster simulator, recreating an Apple website from scratch, completing a scavenger hunt across documents and spreadsheets, fixing bugs gathered from about 15 open-source GitHub repositories, and constructing a simulated bridge that could repeatedly support a car and increasing amounts of weight.

Q: How did Opus 5 perform in the roller coaster test?

Opus 5 created a detailed roller coaster environment with a complete track, trees, buildings, sky, clouds, and a working ride shown from the user’s perspective. The creator says the task finished quickly, consumed about 32,000 tokens, and cost $0.82 in credits. Fable 5 used fewer tokens but cost about 50 percent more and produced less detail.

Q: How did the models compare when recreating the Apple website?

Opus 5 produced the closer recreation according to the creator, although he acknowledged that its output was not perfect. It built recognizable approximations of the page’s cards, people, computers, phones, tablet, watch, and television elements from scratch. Fable 5’s version had cut-off devices, inaccurate product shapes, less convincing people, and more carelessly constructed cards.

Q: What happened during the agentic document test?

The agentic test required each model to search through documents, PDFs, Excel spreadsheets, and other files for eight scavenger-hunt items. Opus 5 located five of the eight items. Fable 5 stopped halfway through after an Anthropic content restriction incorrectly interpreted the task as cybersecurity-related, even though the task only involved locating information in supplied documents.

Q: Why does the creator dislike the personality of Opus 5?

The creator finds Opus 5 excessively verbose, complicated, and prone to moving in many directions instead of remaining focused. He reports stopping responses and asking the model to speak simply and concisely. He also says the behavior can interfere with coding because the model may attempt numerous additional actions before completing a direct request such as fixing a bug.

Q: What are the three main weaknesses of Claude Opus 5?

The three identified weaknesses are its verbose and unfocused personality, Claude’s relatively restrictive overall usage limits, and a Claude Code harness that the creator considers weaker than the ChatGPT app. He also finds Claude’s voice mode substantially less capable than ChatGPT’s voice mode. These limitations prevent him from using Opus 5 throughout his entire workflow.

Q: When should someone use Opus 5, Fable 5, or ChatGPT 5.6?

The creator recommends Opus 5 for especially difficult coding problems because he considers it the smartest of the three models. He retains Fable 5 for extensive planning, brainstorming, conversation, and iterative discussion because its output feels more focused. ChatGPT 5.6 remains his daily driver because he perceives its usage limits as much higher and prefers its app and voice mode.

Summary & Key Takeaways

• Claude Opus 5 won most of the creator’s five practical comparisons against Fable 5. It produced a more detailed roller coaster simulation and Apple website recreation, completed the document scavenger hunt, and fixed open-source bugs more cheaply. Fable 5 won only the bridge test by supporting slightly more simulated weight.

• The creator identifies three major disadvantages of Opus 5: its responses are verbose and unfocused, Claude’s overall usage limits feel restrictive compared with ChatGPT, and the Claude Code harness is less capable than the ChatGPT app. He recommends changing the model’s personality instructions to demand simple, concise, focused responses.

• The creator assigns each model a different role. Opus 5 handles especially difficult coding problems because he considers it the most intelligent option. Fable 5 remains his choice for extensive planning, brainstorming, and conversation because its responses feel more focused. ChatGPT 5.6 is his daily driver because of higher perceived limits and stronger voice features.