METR-fit extrapolation implies 66-hour task horizons for 6.2 Astra, 128 hours for internal model
eliebakouch · x · 2026-11-23
Fitting the latest model results to METR's task-horizon data implies a time horizon of roughly 66 hours for 6.2 Astra and about 128 hours for an internal model, according to an extrapolation shared by @anpaure and amplified by eliebakouch. The estimate follows METR's finding that model task horizons double on a regular cadence, suggesting frontier models' ability to autonomously complete long tasks is climbing rapidly — though it remains an unofficial back-of-envelope fit rather than a benchmark result.
More from AGI Musings
- 'We were here first': an argument that humanity deserves a branch even in an ASI future — ZeroStateReflex · 2026-11-23
- Writer Predicts LLMs Will Train on Bot Output as Human Writing Dwindles — TuhinChakr · 2026-11-23
- Apple's appenz: agent swarms are filling in NVIDIA's CUDA moat — appenz · 2026-11-23
- What comes after know-how? LeCun weighs in on AI redefining intelligence — alexbilz · 2026-11-23
- Uber burned its annual AI budget in 4 months with no visible improvement — yaroslavvb · 2026-11-23
- Post-training may exceed 75% of FLOPs, undercutting perplexity-based safety bars — stochasticchasm · 2026-11-23