The AI labs keep shipping smarter models. Developers, increasingly, are not rushing to keep up.
OpenAI and Anthropic continue to push out new releases, with GPT-6 variants and further Claude iterations planned through mid-2026. Yet a cluster of engagement metrics points to developer growth that is either flattening or migrating toward cheaper options and internal tools.
What the numbers show
Start with SDK downloads, the software kits developers install to plug a model into their own apps. OpenAI’s SDK is pulling in approximately 85.8 million downloads a month, according to 2026 data. Anthropic’s sits at 72.7 million.
Coding agents are where Anthropic has built its strongest position. Claude Code accounts for roughly 81% of public coding-agent commits on GitHub. For comparison, OpenAI Codex has logged about 5.1 million pull requests over its lifetime.
Growth for Claude Code has recently slowed to single digits in annual recurring revenue, the metric software companies use to measure predictable yearly income.
Documentation traffic and the spending squeeze
AI assistant traffic to documentation sites held at about 39% between July and October 2026. Underneath that steady figure, the mix shifted sharply. Traffic driven by ChatGPT dropped by roughly 70% after August 21, 2026, while traffic tied to Claude’s functionality has been rising.
The Ramp AI Index shows an overall decline in AI spending. The drop is attributed to price cuts from OpenAI and Anthropic and to intensifying competition.
Open-source models make up under 5% of business expenditure, according to the Ramp data.
Big tech goes in-house
Meta’s usage of Claude Code fell from about 60,000 users to around 30,000 earlier in 2026. Microsoft, meanwhile, cut its projected spend on Anthropic by more than one third. Both companies are reportedly leaning toward proprietary solutions instead.
What this means for the AI market
There is a competitive wrinkle worth watching between the two labs themselves. Anthropic holds a commanding share of coding-agent activity and is gaining documentation traffic as ChatGPT’s referrals fall. Yet its SDK momentum is cooling, its coding ARR growth has slowed, and two of the largest tech companies have trimmed their reliance on it.
OpenAI keeps the lead in SDK downloads, but the steep drop in ChatGPT-driven documentation traffic suggests developers are spending time elsewhere when they actually sit down to build.
Key metrics to track from here include whether SDK download growth resumes, whether other enterprises copy the move toward in-house tooling, and whether the Ramp index stabilizes once the latest round of price cuts works through the system.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.