Quick Take

Hy3-preview (Tencent) and Claude Fable 5 (Anthropic) represent two distinct approaches to AI deployment. Released in April 2026, Hy3-preview focuses on rapid, affordable reasoning. Claude Fable 5, released in June 2026, prioritizes maximum intelligence and specialized coding performance, albeit at a significantly higher price.

Benchmark Read

Claude Fable 5 consistently outperforms Hy3-preview across all reported benchmarks. With an Intelligence index of 59.9 compared to Hy3-preview’s 33.6, Claude Fable 5 demonstrates stronger capabilities in complex reasoning. Specifically, Claude Fable 5 leads in GPQA (0.926 vs 0.867), HLE (0.533 vs 0.255), and TerminalBench Hard (0.628 vs 0.340). Claude also shows a marked advantage in coding-related tasks, boasting a 76.5 coding index.

Cost and Speed

There is a stark contrast in economic and performance profiles. Hy3-preview is highly economical, with a blended pricing of $0.20/1M tokens, whereas Claude Fable 5 is priced at $20.00/1M tokens.

Performance metrics further differentiate the two. Hy3-preview delivers a rapid output speed of 153.661 tok/s with a time-to-first-token of 1.633s. Conversely, Claude Fable 5 operates at 68.605 tok/s and experiences a significantly longer time-to-first-token of 73.669s, likely due to its adaptive reasoning and multi-stage fallback architecture.

Best Fit

Hy3-preview is best suited for high-volume, latency-sensitive applications where cost efficiency is paramount. It is an ideal candidate for tasks requiring quick, iterative reasoning. Claude Fable 5 is best suited for complex, high-stakes development and analytical tasks where accuracy and intelligence are the primary drivers, justifying its premium cost.