AuraTracer智迹闻
中文

EVENT DOSSIER

[AINews] Claude Fable/Mythos 5.1: new SOTA model, 75% cache price cut but 70% more output tokens

2026-09-02 15:46 Models 🔥 36.0 heat score techmeme #6
1sources
1days unfolding
36.0heat score
3mentions
SummaryAI generated

On September 2, 2026, Anthropic officially released the Claude Fable/Mythos 5.1 model. This version is positioned as “the world’s most advanced programming and knowledge work model,” with Fable 5.1 focusing on autonomous multi-step task execution. In terms of pricing strategy, while the unit prices for inputs and outputs remain unchanged, the cache retrieval price was reduced by 75% to $0.25 per million tokens. However, due to a 1.7-fold increase in average token usage for single tasks, the overall task cost increased by 20% after comprehensive calculations. Community analysis suggests that this version may be based on the same underlying weights as before, with major adjustments made to security policies and routing behavior. The official statement emphasizes that the new model supports autonomous long-term tasks, improves the failure reporting mechanism, and provides enterprise-level zero-data retention control features.

Related eventsRELATED EVENTS
Key entitiesKEY ENTITIES
AnthropicClaudeFable/Mythos 5.1

Coverage · reports per dayLANGUAGE SPLIT

Entity relations
Anthropic × Claude1Anthropic × Fable/Mytho…1Claude × Fable/Mythos 5…1

SignalsSIGNALS

Keyword heat
  • Claude1
  • Anthropic1
  • Fable/Mythos 5.11

All reports (1)SOURCES

L Latent Space en 2026-09-02 15:46

[AINews] Claude Fable/Mythos 5.1: new SOTA model, 75% cache price cut but 70% more output tokens

Anthropic released Claude Fable/Mythos 5.1, positioning it as “the world’s most advanced programming and knowledge work model”. Fable 5.1 focuses on autonomous multi-step tasks. The core pricing remains unchanged for input/output units, but the cache retrieval price was reduced by 75% to $0.25 per million tokens; although manual analysis shows a 1.7-fold increase in token usage for single-task outputs, the overall task cost still increased by 20%. Community analysis suggests that this version may be based on the same underlying weights, with only adjustments made to security and routing behaviors. The official statement emphasizes its support for autonomous long-term tasks, improved failure reporting, and enterprise-level zero-data retention control.