Astra burns tokens at high effort for no gains, finds dev testing Terminal Bench 4.0
zainhas · x · 2026-09-05
Developer zainhas reports that Astra behaves oddly on agentic benchmarks: at high effort it burns through tokens "to no avail."
- Only the low, medium and high effort settings are usable in practice, per the author
- The same pattern reproduces on Terminal Bench 4.0 and deepswe
- The finding points to a mismatch between Astra's effort levels and actual token efficiency
More from Models
- antirez: Astra is a big jump for software dev as RLVR-scaled models keep growing — antirez · 2026-09-05
- Heavy user: ChatGPT's Astra intuitively outclasses Claude Fable 5.1 at the same price — DynaBeast · 2026-09-05
- Fable 5.1 refuses questions on Evoscale/Biohub papers, drawing 'Opus'ed' quip — nathanbenaich · 2026-09-05
- GPT-6 Astra and Claude give opposite answers to the same simple PC check — Jardani_xx · 2026-09-05
- xAI puts Grok 4.6 on its API with a claimed 500K-token context window — emmanuelvivier · 2026-09-05
- Anthropic unveils Enterprise Frontier Safeguards for Claude — emmanuelvivier · 2026-09-05