Loading...
Loading...
GitHub moved Copilot from request-based pricing to a usage-based credit system, mapping credits to $0.01 and selling tiered monthly bundles. Users report sticker shock as routine chats and coding sessions rapidly deplete credits—sometimes exhausting a month’s allotment in a day—with single prompts consuming hundreds or thousands of credits depending on the LLM selected. Critics warn Copilot’s Auto model can pick costly backends for simple tasks, creating unpredictable bills and exposing previously subsidized compute costs. GitHub says the change reflects rising inference expenses and fairness across session types, but the shift could reshape developer workflows, team budgets, and AI tool adoption.
Tech teams rely on predictable developer tooling costs and consistent workflows; sudden usage billing can disrupt budgets and tool adoption. Developers and procurement must reassess cost controls, monitoring, and integration patterns for LLM-driven features.
Dossier last updated: 2026-06-02 04:34:38
Microsoft’s GitHub Copilot switched from flat monthly pricing to usage-based billing, and developers are outraged after reports of rapid credit depletion and surprise charges. Users on GitHub forums and Reddit say hours of routine work consumed large percentages of monthly Copilot Pro+ credits — one reported 16% used in two hours, another said a single request cost more than $6 — prompting threats to migrate to rivals like Anthropic, OpenAI, and open alternatives such as RooCode, LM Studio, and OpenRouter. Microsoft framed the change as aligning cost with heavier, agentic workloads and sustaining the product, but developers say unpredictability hurts budgeting and productivity and are exploring workarounds or platform exits. This could shift developer tool economics and competition in AI-assisted coding.
GitHub has switched Copilot from request-based billing to a usage-based credits model, and users are reporting sticker shock as routine prompts quickly deplete monthly allotments. Under the new plan, credits map to $0.01 each and subscription tiers provide different monthly credit bundles (e.g., Pro $10 = 1,500 credits; Max $100 = 20,000 credits). Costs vary by model and token usage: cheap models like GPT-5.4 nano can be far less expensive than frontier models such as GPT-5.5, and Copilot’s Auto mode can select pricier models unexpectedly. Users shared examples of single prompts or short sessions burning hundreds to thousands of credits, revealing that GitHub had previously been subsidizing heavy inference costs for power users.
GitHub has shifted Copilot from a flat subscription to a usage-based pricing model, prompting strong user backlash on forums and social media. Developers report sticker shock as token- or request-based billing can drive costs far above previous monthly fees for heavy coding, pair-programming, or enterprise integrations. GitHub argues the change aligns charges with actual use and supports sustainability and advanced features, but critics warn it complicates budgeting for teams, penalizes experimentation, and may advantage large firms that can absorb variable bills. The debate matters because Copilot is widely embedded in developer workflows; pricing shifts could affect developer productivity, tool adoption, and competition among AI coding assistants. Expect closer scrutiny from enterprises and potential tooling or policy responses.
GitHub has switched Copilot from request-based billing to a usage-based credit model, and users are reporting rapid depletion of monthly credits and unexpected high costs as the new pricing takes effect. Under the new plan, credits map to $0.01 each and subscription tiers provide different monthly credit bundles (Pro $10 = 1,500 credits; Pro+ $39 = 7,000; Max $100 = 20,000). Cost per prompt varies by token usage and the underlying LLM chosen — cheaper models like GPT-5.4 nano can be far less expensive than frontier models such as GPT-5.5 — and Auto mode may select costly models unexpectedly. Reports show single prompts or short sessions burning hundreds to thousands of credits, revealing how heavily GitHub subsidized heavy users before the change and raising concerns about cost predictability for developers and teams.
GitHub has switched Copilot from request-based billing to a usage-based credits system, and users report sticker shock as routine sessions quickly exhaust monthly allotments. Under the new model, credits map to $0.01 each and monthly plans provide tiered credit bundles ($15–$200 equivalent), but consumption depends on input/output tokens and the LLM chosen — cheaper models like GPT-5.4 nano use far fewer credits than frontier models like GPT-5.5. Users shared examples of single prompts consuming hundreds to thousands of credits, and warn that Copilot’s Auto model can select expensive backends for simple queries. The change exposes how much GitHub previously subsidized heavy Copilot usage and creates cost unpredictability for developers and teams.
GitHub has switched Copilot from request-based billing to a usage-based credit model, and as the new pricing takes effect many users report surprise at steep costs. Users on social media and forums say routine chats and coding sessions now consume large portions — sometimes an entire month’s allotment in a day — leading to projected bills in the thousands compared with prior tiers. GitHub defended the change, saying the old request system equated brief chats and multi-hour autonomous sessions and forced Copilot to absorb rising inference costs. The shift highlights tensions between AI compute expenses and subscription economics, and could affect developer workflows, tool adoption and expectations around AI-assisted coding costs.