chiprook
← AI
AIOctober 16, 2026, 20:18

Ultramarine 1 Flash beats Claude Fable 5 on four of six benchmarks at $0.13 per task

San Francisco startup Paper Instruments unveiled Ultramarine 1 Flash, a model post-trained for $300,000 on top of Z.ai's open-weight GLM-5.3 Flash. It outperformed Claude Fable 5 on four of six knowledge-work benchmarks at roughly one-sixty-third of Fable 5's cost per task, and scored 62.9 on UW JobBench, 17 points above the prior top result. Open weights for independent verification are expected within two weeks of the October 14 launch.

Ultramarine 1 Flash beats Claude Fable 5 on four of six benchmarks at $0.13 per task
#PaperInstruments#Ultramarine#Z.ai#Anthropic
Read next
AI

Z.ai launches GLM-Legal: 19.17% on Harvey LAB at 1/128 the cost of Claude Fable 5

AI

Real-SWE: Fable 5.1 beats GPT-6 Astra and Gemini 3.8 Flash on coding

AI

Test: Meta's Muse Wonder nearly matches Claude Fable 5.1 on agentic tasks

AI

Claude Fable 5: record benchmarks and silent downgrade on AI research queries