Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, with Fable 5.2 featuring enhanced safeguards and a 20% price reduction on API usage. The new models boast a 1 million token context window and improved performance on various benchmarks, including coding and agentic tasks, though Gemini 4 Carbon still leads in business workflow automation. Anthropic also highlighted Fable 5.2's agents successfully formalizing Fermat's Last Theorem in the Lean proof assistant, demonstrating potential for long-horizon, multi-agent work. AI
IMPACT Sets new SOTA on coding benchmarks and demonstrates potential for long-horizon agentic tasks.
RANK_REASON Frontier-lab model release with system card [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Anthropic
- Claude Enterprise
- Claude Opus 5.5
- GPT -6 Astra
- Inkling 2 Small
- Meta
- Muse Ember
- Thinking Machines
- TPU 8i Max
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →