In the first 7 days after Gemini 3.5 Pro's public release, what will Gemini 3.5 Pro's share (%) of tokens in OpenRouter's 'Programming' category be?

Gemini 3.5 Pro's share of tokens in OpenRouter's Programming category during its first seven days post-release is expected to be modest, with a median estimate of roughly 2.5%. This assessment is driven by the specific mechanics of the resolution metric and the structural flow of Google's developer ecosystem. First, the metric-share of tokens-structurally favors cheap, verbose, open-weight models over expensive flagships. As of mid-June 2026, the OpenRouter Programming category is led by models like MiMo-V2.5 (21.5%), MiniMax M3 (11.8%), and Hy3 (11.7%) openrouter.aiopenrouter.aiopenrouter.ai. Even Anthropic, which dominates the category by total spend, sees its top model (Claude Opus 4.7) capture only 4.8% of tokens openrouter.ai. Second, Google's coding usage is heavily undercounted on OpenRouter. The vast majority of Gemini's programming volume flows directly through first-party channels like Vertex AI, AI Studio, and Google's own CLI tools openrouter.ai. Notably, the transition of the Gemini CLI to the Antigravity CLI on June 18, 2026, actively diverts even more individual developer traffic away from OpenRouter developers.googleblog.com. Currently, no Gemini model appears in the OpenRouter Programming top nine, placing the existing Gemini flagship's token share well under the 2.7% cutoff (likely around 1-2%) openrouter.aiopenrouter.ai. While new model launches reliably generate a 7-day testing spike, premium closed models like Gemini Pro have historically struggled to capture dominant token share on OpenRouter against open-weight incumbents. The median estimate of 2.53% assumes a noticeable but constrained launch surge. The distribution includes a long right tail (p90 of 8.1%) to account for the possibility that Gemini 3.5 Pro delivers a generational leap in coding performance, leading to rapid integration into dominant coding-agent workflows and a massive hype-driven spike. Conversely, the downside tail (p10 of 0.57%) reflects scenarios where the OpenRouter listing is delayed, API pricing discourages high-volume token generation, or first-party tools completely cannibalize early OpenRouter adoption.