The Anthropic Claude Fable 5 model confidently outperformed competitors in a realistic physics test, but demonstrated a serious price gap. Its execution cost turned out to be six times higher than that of the more modest, yet significantly more economical Claude Opus 4.8.
Specialists from the atomic.chat service conducted a comparative analysis of four leading AI models, asking them to create three independent HTML5 scenes demonstrating physical processes. The tasks were not easy: a train falling from a destroyed bridge into water, two cars colliding mid-air after jumping from a ramp over a canyon, and a monster truck crushing a row of parked cars. Each model had to correctly render the physics of falling, destruction, and collision, avoiding visual artifacts.
Test Results: Quality vs. Price
Claude Fable 5 became the absolute leader, completing all three scenes with top scores. The collisions looked realistic, objects fell and broke correctly, without falling through the ground or hanging in the air. However, quality came at a cost: the model used 62,158 tokens, with a final price of $3.12.
The closest competitor was GPT-5.5, which used 37,753 tokens and cost $1.14. In the monster truck scene, it even performed slightly better than the rival. Anthropic's Opus 4.8 model proved significantly more economical: 22,280 tokens and just $0.56 — roughly six times cheaper than Fable 5, albeit with more modest rendering quality. The cheapest was GLM-5.2: it didn't win any scene, but cost only $0.08 with 36,246 tokens.
Key takeaway: Fable 5 is the best choice for quality, but you have to pay multiple times more for it.
Context: Fable 5's Return After Restrictions Were Lifted
This test was conducted shortly after Fable 5 became fully available again. As a reminder, on June 30, U.S. authorities lifted export restrictions on the Anthropic Claude Fable 5 and Mythos 5 models, with global access restoration starting on July 1. The restrictions were imposed on June 12, just three days after the models' release, citing national security concerns. In response, Anthropic was forced to disable both models for clients on Claude.ai, via API, AWS Bedrock, and other partner services, as filtering users by nationality in real time proved impossible.
Fable 5 is positioned as Anthropic's most advanced and widely available model. It is built on the Mythos-class architecture with added fine-tuned safety mechanisms and excels at complex tasks, programming, and image analysis. Internal classifiers automatically redirect high-risk queries to safer algorithms.
My opinion: The AI market is clearly moving toward segmentation: there are "premium" models for tasks where quality is critical, and "economical" solutions for mass use. Fable 5 proves that leadership in complex scenarios comes at a cost, and that's normal. However, for 90% of everyday tasks, a sixfold price difference is hardly justified. The choice now is not just a matter of technology, but a matter of economic strategy.