Hands-on with GPT-6.2: Sol shines as daily driver, Astra overpriced with safety quirks
soumitrashukla9 · x · 2026-11-18
Peter Gostev shares real-world (non-demo) experience with GPT-6.2-Astra and GPT-6.2-Sol:
- Sol: his pick as the daily driver — handled data analysis, code migration, and app improvements without issues; solid but not a 2x output jump.
- Astra: expensive (though cache price is down 60%) and quirky — some issues look like RL artifacts, others side effects of safety overreach; he doubts it's worth the price.
- Notable weakness: poor at interpersonal/theory-of-mind tasks (e.g., submitting an IT ticket), frequently getting blocked or hesitating over random things, which he attributes to safety overcorrection.
More from Models
- Domingos on iWSP paper: just use the program instead of writing it into LLM weights — pmddomingos · 2026-11-18
- GPT-6.2 Astra and Sol post highest error rates ever on VoxelBench 2 — legit_api · 2026-11-18
- GPT-6.2 Astra hits new harmonic milestones on the Bach Benchmark — deanwball · 2026-11-18
- Early hands-on: GPT-6.2 wins many benchmarks and is surprisingly cheap — ramos_casals · 2026-11-18
- Chutes partners with Alibaba Cloud to offer GLM-5.3 and DeepSeek V4.1 at 35% off — markjeffrey · 2026-11-18
- Meta launches Muse Spark Open: 816B MoE coding model with 42B active params and 1M token context — PyTorch · 2026-11-18