Anthropic Releases Claude Fable 5, a Safer Alternative to Mythos
Anthropic has introduced Fable 5, an AI model considered 'safer' than the unreleased Mythos, which poses cybersecurity concerns and is limited to a few enterprises. Fable 5 demonstrates notable advancements over Opus 4.8, although it lags behind GPT 5.5. It is capable of functioning for long durations and generating multiple subagents while maintaining context. Initially accessible through Claude subscription plans until June 22, it will transition to a paid credit system afterward. Fable is priced at twice that of Opus, making it more affordable than Mythos, which costs five times Opus. The launch has sparked controversy due to a policy aimed at 'secretly' undermining ML/AI projects. In other developments, Google has released Gemini 3.5 for real-time translation, and there are updates regarding Claude Code and ChatGPT's model selector.
Key facts
- Anthropic released Fable 5, a safer version of Mythos.
- Mythos is unreleased and considered a big cybersecurity risk by Anthropic.
- Fable 5 shows a big benchmark jump over Opus 4.8, but not vs GPT 5.5.
- Fable 5 can work longer and spawn dozens of subagents reliably.
- Fable 5 is available in Claude subscriptions until June 22, then moves to paid credits.
- Fable 5 is 2x more expensive than Opus; Mythos was 5x Opus.
- Anthropic introduced a policy to secretly sabotage ML/AI work, later partially walked back.
- Google launched Gemini 3.5 Live Translate for real-time speech-to-speech in 70+ languages.
Entities
Artists
- Ethan Mollick
- Dan Shipper
- Dario Amodei
- Keshav
- Ben Tossell
Institutions
- Anthropic
- OpenAI
- GitHub Copilot
- Replit
- Factory
- Intercom
- Orb
- Cursor
- Plaid
- SemiAnalysis
- Firecrawl
- Supermemory
- Skribe
- Mainframe
- FluidInference