GPT-5.6 Sol Reportedly Hit by Reasoning Budget Cut
johnseach · x · 2026-07-13
A post claims that shortly after release, GPT-5.6 Sol became "faster but shallower": solving hard problems on day one, but later requiring higher settings or simply performing worse.
The author attributes this to a massive cut in the internal reasoning budget ("juice"): max juice dropped from around 960 to 128, affecting xHigh, High, and lower tiers, while cheaper models like Terra and Luna remained unchanged. OpenAI allegedly explained this as increased usage from reasoning optimizations, a rollback of context due to runaway costs, and stated that related juice experiments have been "reverted to normal".
The author concludes that:
- Peak intelligence is becoming an adjustable parameter rather than a fixed attribute
- Pro/Codex users might get more quota, but the model's depth has decreased
- Cheaper models might ironically be the smarter choice for daily use
- A model's state on launch day is unstable, which damages user trust
Related event: GPT-5.6 Sol Silently Nerfed, OpenAI Confirms Rollback(8 posts)→
More from Models
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11
- Looking for a classifier of software engineering task shapes to pick models per task — StewartalsopIII · 2026-09-11
- DeepSeek V4 Pro API to continue after Sept 2026, billing unchanged — teortaxesTex · 2026-09-11
- DeepSeek V4.1 Flash Hits 98% of GPT-6 Astra's Score at 1.4% of the Cost in Third-Party Benchmark — ayushtweetshere · 2026-09-11
- TheZvi Polls: Has Your Coding Model Choice Changed Since Fable 5.1 and Astra? — TheZvi · 2026-09-11
- antirez Weighs In on Anthropic Banning Minors From Using Claude — antirez · 2026-09-11