The latest OpenAI model GPT-5.6 went public on July 9, 2026. Meanwhile, Anthropic has dropped two frontier models: Opus 5 and now Fable 5.1. That's wild, OpenAI might drop a new model soon.
Anthropic
- Claude Opus 5 — July 24
- Claude Fable 5.1 and Mythos 5.1 — Sept 1
That's 2 Claude releases since GPT-5.6. And other labs too haven't been quiet.
Everyone else
- Kimi K3 (Moonshot AI) — July 16
- GLM-5.3-Flash (Z.ai) — August
The rumor: Anthropic might be rushing an Opus 5.1 out alongside Fable 5.1, timed against whatever OpenAI is cooking with Astra.
But let's see the benchmarks of Fable 5.1 (I am yet to test the model on my personal hard tasks, but till then the benchmarks are what we have):
The Numbers
Agentic scientific research (Terminal-Bench-Science 0.1) 52.6% — that's 2.4x GPT-5.6 Sol and 2.1x Fable 5!
Agentic coding (Terminal-Bench 4.0) 55.8%, up from 42 on Fable 5. GPT-5.6 Sol is 37.3. And the footnote is the fun part: 60.9% when you pair it with Mythos 5.1.
Knowledge work (GDPval-AA v2) 1853 Elo. +130 over Fable 5, +29 over Opus 5.
Computer use (OSWorld 2.0) 77.9% partial, 41.7% strict. That strict number is the first time anything crossed 40 on this benchmark.
Reasoning (Humanity's Last Exam) 60.9% without tools, 65% with. Barely a 4 point gap between the two. The base model got smarter.
Business workflows (AutomationBench) 31.4% — nearly double Fable 5, 1.6x GPT-5.6 Sol. Biggest relative jump on the board.
Agentic coding again (CursorBench 3.2.0) 73.4%. Wins here too (+2.9 over Fable 5).
Thoughts
I am yet to test the model, but the more interesting point is OpenAI has been quiet while Anthropic and others keep dropping frontier models.