OpenAI Overtakes Anthropic in OpenRouter Wallet Share for First Time in 2.5 Years

OpenRouter's official data shows that in the week of September 7, user spending (share of wallet) on OpenAI models exceeded that on Anthropic models for the first time in over 30 months. Gavin Baker, citing OpenRouter data, noted that over two months since June, OpenAI's share relative to Anthropic rose from 20% to 50%, while Anthropic fell from 80% to 50%. "Wallet share" has long been viewed as a truer paid-usage signal than call volume, though the data reflects only one aggregation platform and the spending and token-usage metrics have diverged.

Confirmed

• OpenRouter data shows OpenAI overtook Anthropic in user spending last week, the first reversal in about 2.5 years (@firstadopter, @FinanceYF5, @AccBalanced and others relayed the same official data).
• @新智元 pinpointed the reversal to the week of September 7, the first since early 2024—more than 30 months.
• @GavinSBaker cited OpenRouter data showing OpenAI's share rising from 20% to 50% over two months, with Anthropic dropping from 80% to 50%; @FinanceYF5 relayed the same shift.
• @gdb relayed data shared by OpenAI president Greg Brockman from analyst Peter Walker: the top models by user spend and by token consumption differ, showing divergence between the two metrics.
• @AccBalanced added context: Anthropic's coding models long dominated paid usage on the platform, and OpenRouter data is widely seen as a market bellwether.
• mikeldking, relayed by @aparnadhinak, offered one explanation: users are spending heavily testing the newly released Astra, especially on pricier tiers—a transitional consumption pattern.

Unconfirmed

• Posts conflict on which model tops each metric: @gdb's relay says Anthropic's Astra led by spend, while @FinanceYF5's summary says GPT-6 Astra led spending and GPT-5.6 Luna dominated token usage; the original OpenRouter data is needed to reconcile.
• Dr. Eli David's comment that users are moving away from Anthropic is personal interpretation not directly supported by the data; Rohan Paul's "OpenAI is so back" framing is likewise an emotional take.
• The "new model testing" explanation comes from a single relayed comment and lacks further supporting data.

Why it matters

• @firstadopter argues the shift may reflect newer OpenAI models' appeal or cost-effectiveness changing the API landscape; @FinanceYF5 cautions it reflects only OpenRouter's single platform, and the divergence between spend and token metrics suggests both must be watched.
• If the 20%→50% shift holds, it indicates a structural migration in coding/agent model choice over two months rather than a one-week blip; whether Astra-driven testing spend is a short-term factor will be tested by coming weeks' data.
• From a capital-markets view, if Anthropic's slowdown is read as declining competitiveness, it could affect its IPO narrative—one reason the data drew wide attention.

2026-09-16 ~ 2026-09-17 · 13 related posts

• [source] OpenRouter spend flips to OpenAI over Anthropic for first time in 2.5 years — firstadopter · 2026-09-16
• Anthropic's Astra tops spend while OpenAI's Luna dominates token usage by a lot — gdb · 2026-09-16
• OpenRouter users spent more on OpenAI than Anthropic models, first time in 2.5 years — soumitrashukla9 · 2026-09-16
• OpenAI spending overtakes Anthropic on OpenRouter for the first time in 2.5 years — FinanceYF5 · 2026-09-16
• OpenRouter's two leaderboards diverge: GPT-6 Astra tops spending while GPT-5.6 Luna dominates token usage — FinanceYF5 · 2026-09-16
• [source] OpenAI overtakes Anthropic on OpenRouter spending for the first time in 2.5 years — 新智元 · 2026-09-16
• [source] OpenAI's OpenRouter share jumps from 20% to 50% vs Anthropic in two months — GavinSBaker · 2026-09-16

6 near-duplicate retellings: SumitGup · AccBalanced · aparnadhinak · rohanpaul_ai · ivan_bezdomny · FinanceYF5