A Line That Moves Every Few Weeks
The artificial intelligence race is no longer governed by annual product cycles. Its defining threshold now changes within weeks, as every new flagship model attempts to redraw the frontier for reasoning, coding and autonomous agents.
Anthropic raised that threshold with Fable 5 in June. The model reached 60 points on the Artificial Analysis Intelligence Index, establishing itself as the benchmark leader for advanced knowledge work and software engineering.
xAI accepted the challenge by releasing Grok 4.5 on July 8. The model led some engineering evaluations, including SWE Marathon, but did not comprehensively surpass Fable 5. It remained behind Anthropic’s model on Terminal-Bench and SWE-Bench Pro.
OpenAI followed a day later with GPT-5.6 Sol. It did not seize the outright lead, but came within one point of Fable 5 on the Artificial Analysis index while offering a competitive combination of speed, intelligence and cost. The contest is therefore no longer simply about which model is smartest, but which delivers the strongest performance per dollar and per second.
Was Google Really Late?
Chronologically, no. Google released Gemini 3.5 on May 19, before Fable 5, Grok 4.5 and GPT-5.6. Its difficulty is that competitors subsequently moved the performance line.
Gemini 3.5 Flash currently scores about 50 on the Artificial Analysis Intelligence Index, compared with 60 for Fable 5 and 59 for GPT-5.6 Sol. It is considerably faster and cheaper than OpenAI’s flagship, but it no longer occupies the leading position in overall intelligence.
There is no public evidence that Google is withholding another model because it fears poor benchmark results. Its release timing may instead reflect reliability testing, infrastructure capacity, cost management and the challenge of integrating Gemini across Search, Workspace and Android. Claims that Google is delaying because its model cannot cross the line remain speculation.
Why Google Is Still in the Race
- Distribution: Search, Android, YouTube and Workspace give Gemini exceptional global reach.
- Infrastructure: Google owns its TPU architecture and operates one of the world’s largest cloud platforms.
- Efficiency: Gemini 3.5 Flash remains highly competitive on price and speed.
- Developer risk: A prolonged performance gap could push companies to build their products around rival models.
Between the Lines
Google has not fallen out of the competition, but it has temporarily lost control of the industry narrative. Its rivals are releasing the models that define the frontier, leaving Google to appear like the pursuer rather than the pioneer.
What Comes Next?
Google’s next Gemini release will be decisive. If it returns to the top of independent benchmarks while preserving its advantages in cost and distribution, the current gap will look temporary. If it again falls short, Google will face an uncomfortable question: how can a company with world-class researchers, infrastructure and distribution still fail to produce the market’s strongest model?
