GitHub [added Claude Opus 4.7]1 to Copilot Pro+, Business, and Enterprise on April 16, 2026, at a 7.5× premium-request multiplier labeled promotional — expiring April 30. At that rate, a 1,500-request monthly allowance yields roughly 200 Opus 4.7 turns, versus ~500 turns at Opus 4.6’s 3×. The post-promotion multiplier is undisclosed. Simultaneously, Opus 4.5 and 4.6 are being removed from the Pro+ model picker, and billing records in community threads document 7.5× charges on models users claim they never invoked.
The April 16 Launch and the 7.5× Multiplier
Copilot’s premium-request system is GitHub’s own accounting layer sitting on top of the Anthropic API — it is not the same as Anthropic’s per-token pricing, which remains unchanged. Per [GitHub’s documentation on premium requests]2, each model carries a fixed multiplier applied against a plan’s monthly allowance: Opus 4.6 costs 3 premium requests per prompt, Opus 4.7 costs 7.5, and Opus 4.6 in fast mode costs 30. The fast-mode figure shows the multiplier can scale well beyond what Anthropic’s underlying token pricing would suggest — a point community users have already noted.
For a Pro+ plan with 1,500 monthly premium requests, the arithmetic:
| Model | Multiplier | Turns from 1,500 allowance |
|---|---|---|
| Opus 4.6 | 3× | ~500 |
| Opus 4.7 (promotional) | 7.5× | ~200 |
| Opus 4.6 fast mode | 30× | ~50 |
The 7.5× figure is explicitly labeled promotional through April 30. The post-promotion multiplier does not appear in any public documentation as of April 24.
For Business and Enterprise accounts, [the April 16 changelog]1 notes that Opus 4.7 requires administrator enablement — it is not automatically active. Organizations that have not switched the model on are not yet exposed to this pricing shift.
Why Opus 4.6 Was Removed and What That Forces on Pro+ Users
Removing Opus 4.5 and 4.6 from the Pro+ model picker — [announced in the April 20 plan changes]3 — eliminates the fallback option. Previously a developer could select Opus 4.6 at 3× and stay within budget across a long agentic run. Once those models are removed from the picker, the choice collapses: use Opus 4.7 at whatever multiplier survives April 30, or drop to a cheaper model class entirely.
[Community discussion #192814]4 documents users doing their own cost math against direct API pricing differences, estimating that an equivalent API cost differential would land around 4×, not 7.5×. GitHub has not offered a public explanation of how the promotional multiplier was derived. The thread also confirms Opus 4.6 was removed rather than demoted to a lower-cost tier.
Billing Disputes: 7.5× Charges for Models Never Invoked
The more operationally disruptive report is in [community discussion #192911]5, where a user posted billing records showing 7.5× charges attributed to Claude Sonnet 4.6, Claude Haiku 4.5, and Gemini 3 Flash — three models the user states they did not invoke. The charges resulted in roughly an $18 overcharge and exhausted the user’s premium quota ahead of schedule.
This is a single-user report, not an audited billing discrepancy or independently confirmed pattern. But the specifics are concrete: named models, a cited multiplier, documented quota exhaustion. GitHub’s response was to close the discussion and direct the user to open a support ticket, stating the community forum is not a support channel.
The structural issue is not whether the dollar figure is correct — it’s what the incident reveals about observability. A developer running agentic workflows cannot easily audit which model each sub-call resolved to, or what multiplier was applied per invocation. If a background agent resolves to a more expensive model than the user explicitly selected, the cost surface is larger than it appears from the session log.
What Happens After April 30?
[The April 16 changelog]1 labels the 7.5× rate as promotional and states it expires April 30. No follow-up announcement has disclosed the post-promotion multiplier as of April 24.
The narrow decision window — 14 days between launch and promotion expiration — creates a forced choice that cannot be precisely priced. Unlike a standard promotional discount where the post-promo price is stated upfront, the base post-promotion rate is hidden. That’s an unusual structure for enterprise tooling, where procurement workflows typically require forward cost visibility.
The April 20 plan changes add context: [GitHub paused new self-serve signups for Copilot Pro and Pro+]3 and introduced tighter usage limits for individual plans, with Pro+ offering more than 5× the limits of Pro. Existing subscribers are not cut off, but the pause on new signups signals active supply-side management of the individual tier.
The Shift from Flat Subscriptions to Metered Turns
The pattern here isn’t unique to Opus 4.7. Copilot’s premium-request model already embeds variable per-turn costs behind a subscription facade. A developer paying a flat monthly fee for Pro+ is, functionally, buying a finite request budget segmented by model tier. Adding a 7.5× model while removing the 3× alternative is a pricing move that raises the cost floor of a single agentic session without changing the subscription line item on the invoice.
The second-order consequence: prompt cost becomes a variable an individual engineer has to track before committing a turn. In a workflow where an agent triggers sub-calls, refines context, and reruns on failures, a single “task” can consume multiple premium-request slots at compounding multipliers. The #192911 billing incident shows what that looks like when model selection isn’t transparent to the user.
This is what per-seat, opaque-multiplier billing produces at scale: cost management migrates from finance and procurement — who see the subscription total — down to the engineer, who now has to reason about the multiplier before sending a prompt. That’s a different burden allocation than flat subscriptions, and one that compounds as agentic sessions grow longer and more autonomous.
Frequently Asked Questions
Does the 7.5× multiplier affect Copilot Pro subscribers differently than Pro+?
Pro subscribers have roughly 300 premium requests per month (the 1,500 Pro+ allowance divided by five), yielding only ~40 Opus 4.7 turns at 7.5× — barely enough for two sustained agentic sessions. The article’s arithmetic focuses on Pro+, but the 5× smaller Pro allowance makes Opus 4.7 functionally inaccessible on the cheaper tier for any multi-step workflow.
The #192911 billing dispute includes a Google model — does the misattribution cross vendors?
Yes. The 7.5× charges span Anthropic models (Sonnet 4.6, Haiku 4.5) and Google’s Gemini 3 Flash, so the billing misattribution isn’t confined to a single provider’s integration. The observability gap likely sits in Copilot’s model-routing layer — GitHub’s own infrastructure — meaning any model routed through Copilot could be affected, regardless of vendor.
What does Opus 4.6 fast mode’s 30× multiplier reveal about GitHub’s ceiling for these rates?
At 30×, the full Pro+ allowance yields ~50 turns — fewer than two agentic sessions involving context gathering, generation, and retry loops. That GitHub is simultaneously removing fast-mode Opus 4.6 from the picker suggests even it considers that rate unsustainable for normal use. The precedent matters: multipliers well above 7.5× already exist on the current rate card.
What happens if the post-April-30 multiplier lands above 7.5×?
At 15×, the 1,500-credit Pro+ allowance drops to ~100 Opus 4.7 turns; at 30×, to ~50. With the 3× Opus 4.6 fallback removed, there is no mid-tier Opus escape hatch. The concurrent pause on new Pro/Pro+ signups signals active supply management, making a higher post-promo rate plausible. Teams relying on Opus-grade agentic workflows should plan for either additional allowance packs or a migration to direct API access before May 1.