GitHub frames the shift as 'paying for what you use,' arguing that expanding capabilities like multi-file edits, agentic coding via Copilot Workspace, and support for frontier models like Claude 3.5 Sonnet and Gemini consume far more compute than simple inline completions. The company positions the change as aligning cost with value delivered rather than as a price increase.
The editorial argues that flat-rate pricing was the industry's implicit bet that AI completions would get cheap enough to subsidize unlimited usage, and GitHub just admitted that bet doesn't hold when agentic workflows and frontier models are added to the mix. A single agentic task can make dozens of LLM calls behind the scenes, making flat rates unsustainable.
The editorial notes that while GitHub frames this as 'paying for what you use,' the math tells a more complicated story. Each plan now includes a base allocation of premium requests with overage billing, meaning power users who previously had unlimited completions at $10/month will likely face higher costs. The editorial identifies a 'sharp split' in the Hacker News community over whether this framing is honest.
The editorial explicitly calls the more generous free tier 'a clear play to keep the top of the funnel wide while monetizing heavy usage.' This two-tier strategy lets GitHub maintain its massive developer adoption numbers while shifting the cost burden to the enterprise and power-user segments that consume the most expensive model interactions.
The editorial argues this isn't just a pricing change but 'a fundamental shift in how AI-assisted development gets monetized.' For three years the industry assumed AI completions would become cheap enough to offer unlimited usage; GitHub's move away from that model has implications for every competitor offering flat-rate AI coding tools.
GitHub announced that Copilot is moving away from its flat-rate subscription model to usage-based billing. The shift affects all tiers — Free, Pro, Business, and Enterprise — and introduces the concept of "premium requests" for interactions that use more expensive models or agentic capabilities.
Under the old model, individual developers paid $10/month (or $100/year) for Copilot Pro, and organizations paid $19/seat/month for Copilot Business. Everyone got unlimited completions from the base model. Under the new model, each plan includes a base allocation of premium requests, and usage beyond that allocation is billed per request. The free tier actually gets more generous — a clear play to keep the top of the funnel wide while monetizing heavy usage.
The change takes effect for new subscribers immediately, with existing subscribers transitioning over the coming months. GitHub is framing this as "paying for what you use" rather than a price increase, but the math tells a more complicated story.
This isn't just a pricing change — it's a fundamental shift in how AI-assisted development gets monetized. For the past three years, flat-rate pricing was the industry's implicit bet that AI completions would get cheap enough to subsidize unlimited usage. GitHub just admitted that bet doesn't hold when you add agentic workflows and frontier models to the mix.
The timing is telling. Copilot's feature set has expanded dramatically: multi-file edits, agentic coding tasks via Copilot Workspace, support for third-party models like Claude 3.5 Sonnet and Gemini. These capabilities consume significantly more compute than simple inline completions. A single agentic task might make dozens of LLM calls behind the scenes. Flat-rate pricing made that unsustainable.
The Hacker News community response (591 points and climbing) reveals a sharp split. One camp sees this as an inevitable and honest correction — you can't offer unlimited GPT-4o-class completions for $10/month without hemorrhaging money. The other camp sees it as a bait-and-switch: hook developers on unlimited access, wait until Copilot is embedded in their workflow, then turn on the meter. Both camps have a point.
What's particularly notable is the "premium request" distinction. Base completions — the bread-and-butter tab-complete suggestions — appear to remain effectively unlimited. The metering kicks in when you use frontier models or agentic features, creating a two-tier experience where the most powerful capabilities are the ones that cost extra. This mirrors how cloud providers learned to price compute: give away the basics, meter the premium tier.
The competitive implications are immediate. Cursor, Windsurf, Cody, and every other AI coding assistant now face a strategic choice: match GitHub's usage-based model (validating the approach) or double down on flat-rate as a differentiator. Cursor in particular has been aggressive with its $20/month Pro plan that includes significant Claude and GPT-4 usage. If GitHub's move signals that flat-rate is unsustainable, Cursor's pricing is on borrowed time too.
For individual developers, the impact depends entirely on your usage pattern. If you mainly use Copilot for inline completions and occasional chat, your bill likely stays flat or even drops. If you've been leaning heavily on agentic features, multi-file edits with frontier models, or using Copilot as your primary coding interface, budget for a 2-5x increase over your current subscription cost.
For engineering organizations, this changes the procurement conversation significantly. A 50-person team paying $950/month for Copilot Business had predictable costs. Under usage-based billing, that same team's bill becomes variable — and the variance correlates with exactly the behavior you want to encourage (using AI more). CTOs and engineering managers need to:
1. Audit current usage patterns before the transition hits. GitHub's admin dashboard should show premium request volumes per seat. 2. Set spending limits per seat or per team. Usage-based billing without caps is how cloud bills spiral. 3. Evaluate whether the premium models justify the premium cost. For many tasks, the base model completions are good enough. Reserve frontier model usage for complex refactors and architectural work. 4. Benchmark alternatives. This is a natural moment to trial Cursor, Cody, or local models via Continue.dev. If you're going to pay per-request anyway, the switching cost argument for staying with Copilot weakens.
There's also a subtler architectural implication: usage-based pricing creates an incentive to minimize LLM round-trips, which may push teams toward local models for high-frequency, low-complexity completions while reserving cloud models for harder problems. This hybrid approach — local for speed and cost, cloud for capability — may become the default architecture for AI-assisted development.
GitHub's move is a market signal, not just a product update. The era of all-you-can-eat AI coding assistance at $10-20/month is ending. The question isn't whether other tools follow — it's how quickly. Developers who treat this as a wake-up call to understand their actual AI usage patterns, rather than just absorbing a bill increase, will be better positioned regardless of which tool they use. The winners in this next phase won't be the cheapest providers; they'll be the ones whose metering is transparent enough that developers trust the bill.
The era of subsidised inference is truly ending. The new model multipliers (https://docs.github.com/en/copilot/reference/copilot-billing...) seem like a huge leap, though. From 1x to 6x for new-ish GPT and Sonnet models. 27x for Opus...Seems like folks would be better o
"Your plan pricing is unchanged: Copilot Pro remains $10/month and Pro+ remains $39/month, and each includes $10 and $39 in monthly AI Credits, respectively."If there's no discount on credits (in terms of tokens per dollar) over other providers, I'm going to switch to a
Well.Just got an email from GitHub saying they'll be raising prices for Co Pilot."To keep up with the way you use Copilot, we're transitioning to usage-based billing, and we want to give you enough time to prepare."Man, it was fun. Having my tokens subsidized by Microsoft. If the
Everybody who says it's a 5-9-27x seems to not be aware of the obvious loophole. More like 50x increase. You were able to use over $500 worth of Opus on a $10/mo Github plan easily, no hacks. You could just prompt "plan this out for me, don't stop until fully planned, don't
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
Something is hilariously off here: Why should I pay $10 and be forced to use it by the end of the month, while I can pay $10 and have it last as long as I want?Their "API pricing" is exactly the same as that of providers: https://docs.github.com/en/copilot/referenc