Leaked first impressions of GPT 6.2 Astra: smartest yet hardest to communicate with
soumitrashukla9 · x · 2026-11-18
Blogger HistoryGPT (amplified via RT) claims weeks of early access to GPT 6.2 Astra and shares first impressions (unverified by OpenAI — treat as a rumor):
- Smartest but poor communicator: he calls it the smartest model he's worked with in raw intelligence, but like many very smart humans it struggles to communicate effectively.
- Excels at long-running autonomous knowledge work: indexing thousands of pages, hunting sources in obscure places, managing other agents for actual research — it figures out the full scope, works laterally, and emulates the best human research work in his field.
- He says he made new historical discoveries with it (a first for him with an LLM), with details to come; he rates it above Fable and Opus for such tasks.
Other developers report the opposite experience — benchmarks saturated but demonstrably worse than Fable and Opus 4.5 in personal testing — so real-world reception is sharply split.
More from Models
- Google ships Gemini 4 Hydrogen, a faster robotics model, plus Antigravity Robotics plugin — keerthanpg · 2026-11-18
- Epoch: GPT-6.2 Astra scores 174 on ECI, above the pre-Astra reasoning-era trendline — airesearch12 · 2026-11-18
- Claude delegates synthetic data generation to local Qwen3.8-27B, citing its own guardrails — maksym_andr · 2026-11-18
- GPT-6.2 stuck in compaction loop in Codex, burning 4.44M tokens in 9 minutes — Dear_Lock_5280 · 2026-11-18
- Inkling-2-Small runs on two DGX Spark units and is claimed competitive with DSv4.1 and GLM 5.3 — LiTianleli · 2026-11-18
- Users find GPT-6.2 keeps setting its generated games in the same fictional place — kieranklaassen · 2026-11-18