Anthropic's new flagship model, Claude Fable 5, confidently won a realistic physics test, demonstrating impressive simulation of complex scenes. However, the victory came at a cost: according to independent testing, the cost of completing tasks was approximately six times higher than its predecessor, Opus 4.8.

The comparison took place on the atomic.chat platform, where four leading models were tasked with creating three independent HTML5 scenes demonstrating physical processes. The scenarios included: a train derailing from a destroyed bridge into water, two cars colliding mid-air after a ramp jump over a canyon, and a monster truck crushing a row of parked cars. Each model had to accurately depict the physics of falling, destruction, and collisions without extraneous artifacts.

Test Results: Fable 5 Unrivaled, but with a Caveat

Claude Fable 5 handled the task flawlessly, completing all three scenes with top marks. Collisions looked realistic, objects fell and broke correctly, without clipping through the ground or hanging in the air. However, quality came at a price: the model consumed 62,158 tokens, with an execution cost of $3.12.

The closest competitor was GPT-5.5, which consumed 37,753 tokens at a cost of $1.14. In the monster truck scene, this model even performed slightly better. Opus 4.8 proved significantly more economical — 22,280 tokens and only $0.56. This is roughly six times cheaper than Fable 5, but the quality of scene rendering was considerably more modest. The cheapest was GLM-5.2, which didn't win any scenes but cost only $0.08 with 36,246 tokens.

My Analysis: The choice between Fable 5 and Opus 4.8 is a classic "quality versus price" dilemma. For tasks where the accuracy of physical simulation is critical, Fable 5 is the undisputed leader. But for mass-market or budget projects, Opus 4.8 remains a more rational choice, and the price gap here is enormous. The AI market is clearly moving toward segmentation: premium models for complex tasks and economical versions for routine work.

Fable 5's Return After US Restrictions Lifted

The test took place shortly after Fable 5 became fully available again. As a reminder, on June 30, US authorities lifted export restrictions on Anthropic's Claude Fable 5 and Mythos 5 models, with global access restoration beginning on July 1. The restrictions were imposed on June 12, just three days after the models' release, citing national security concerns. 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, built on the Mythos-class architecture with finely tuned safety mechanisms. It excels in complex tasks, programming, and image analysis. Prices remain competitive: $10 per 1 million input tokens and $50 per 1 million output tokens. This cost was reflected in the atomic.chat test.

My Conclusion: Fable 5's victory in the physics test is not just a marketing success but real proof of progress in simulating complex environments. However, the six-fold price gap with Opus 4.8 raises the question: is the market ready to pay such a premium for realism? For developers of games, simulators, and VR content — undoubtedly. For others — it remains a question of economic efficiency.