Claude Fable 5
PAIDAnthropic's most powerful generally-available model in the Mythos class. Released June 9, 2026, Fable 5 scores 95.5% on SWE-bench Verified and 80.3% on SWE-Bench Pro, leading all frontier models by a wide margin in coding, long-horizon agentic tasks, and vision reasoning.
📊 Ratings
4.8/ 5
4.9
Features
4.5
Ease of Use
4.0
Value
4.3
Support
✅ Pros
- ✓State-of-the-art on nearly all tested benchmarks
- ✓11-point lead over Opus 4.8 on SWE-Bench Pro
- ✓Exceptional long-horizon agentic coding capability
- ✓Stripe completed 50M-line Ruby migration in 1 day (vs 2+ months estimated)
- ✓Advanced vision: rebuilt web apps from screenshots, completed Pokemon FireRed from vision alone
- ✓1M token context window for massive codebases and documents
- ✓Self-verification and autonomous task execution
❌ Cons
- ✗Double the price of Opus 4.8 ($10/$50 per MTok vs $5/$25)
- ✗Mandatory 30-day data retention with no zero-retention option
- ✗Safety classifiers silently reroute ~5% of sessions to Opus 4.8
- ✗Currently suspended by US government export directive (as of June 12, 2026)
- ✗Adaptive thinking always on, cannot be disabled
- ✗Raw chain-of-thought never returned to API users
- ✗Overly aggressive safety filters on biology/chemistry topics
Comments (0)
Join the conversation
Log in to commentNo comments yet. Be the first to share your thoughts!