Fable 5 vs GPT-5.6 vs Gemini 3.5 Pro: Three Flagship Models in One Month - Who Actually Wins?

[Sharp Shannon]

June has been a genuinely strange month for frontier AI because three major model releases landed almost simultaneously. Anthropic launched Fable 5 on 9 June, topping SWE-bench Verified at 95 percent for coding and posting the largest margin improvement in Agent Arena history at 11.2 percent net overall gain over the previous leaderboard leaders. Then OpenAI launched GPT-5.6, codenamed Kindle, with claimed improvements in reasoning, coding, vision and agentic workflows. Google's Gemini 3.5 Pro is competing in the same capability tier. All three are going after the same enterprise customers, the same developer workflows, and the same benchmark headlines in the same four-week window.

The commercial context makes this more interesting than usual benchmark wars. Anthropic has just overtaken OpenAI in US business adoption for the first time, with 34.4 percent of business spend versus OpenAI's 32.3 percent. Anthropic is deriving about 80 percent of revenue from enterprise against OpenAI's 40 percent and generates roughly 16.20 dollars in monthly revenue per active user against OpenAI's 2.20 dollars. OpenAI has seven times more users but Anthropic is making more money per user by a factor of seven. GPT-5.6 reportedly comes with aggressive pricing cuts to try to close that enterprise gap. The three-way race is now genuinely open in a way it has not been for the past 18 months when Anthropic was consistently ahead on coding benchmarks.

Ignoring the benchmark numbers for a moment: which model are you actually using day to day, has that changed recently, and what would it take to make you switch?

[CMPunk_Fan]

The revenue per user gap is the number that actually matters here. Anthropic has fewer users but each one is worth seven times more. That is a structurally better business even if OpenAI wins on headline user count