GitHub Copilot's Claude Opus multiplier jumps to 27x on June 1. Monthly plans dodge the hike.

GitHub published a new model multiplier table for Copilot Pro and Pro+ annual subscribers, and the numbers are jarring. Claude Opus 4.6 goes from 3 to 27 per request. Sonnet 4.6 goes 1 to 9. Gemini 3 Pro goes 1 to 6. Annual plans on request-based billing absorb the jump June 1.

The catch, and it’s the part the docs page buries near the top, is that the new multipliers only apply if a subscriber stays on a request-based annual plan. Move to monthly, or accept the broader usage-based billing flip GitHub announced on April 27, and the table doesn’t apply. “Starting June 1, your Copilot usage will consume GitHub AI Credits,” GitHub wrote in the announcement, but only if a subscriber accepts the new model. The structure penalizes the customers who paid up front for a year of stable pricing.

What’s actually changing

Each “premium request” today represents one prompt to a non-default model. A monthly bucket ships with each plan, and a multiplier scales how fast the bucket drains. A 27x multiplier on Opus 4.6 means a single chat now spends 27 requests instead of 3.

The headline rows from GitHub’s docs:

• Claude Opus 4.6: 3 → 27 (9x more expensive)
• Claude Opus 4.7: 15 → 27 (already bumped from 7.5 in April)
• Claude Opus 4.5: 3 → 15
• Claude Sonnet 4.6: 1 → 9
• Claude Sonnet 4.5: 1 → 6
• Gemini 3 Pro / 3.1 Pro: 1 → 6
• GPT-5.4 and GPT-5.3-Codex: 1 → 6
• GPT-5.4 mini: 0.33 → 6, an 18x jump on what was billed as a cheap model
• GPT-5.1, GPT-5.2, GPT-5.1-Codex, GPT-5.1-Codex-Max: 1 → 3

A few models stay flat: Sonnet 4, Gemini 2.5 Pro, Haiku 4.5, Gemini 3 Flash. Older OpenAI models that were free under the current plan (GPT-4o, GPT-4.1, GPT-5 mini) start charging 0.33 to 1 per request. Even those flat rows are a tax for users on annual plans who chose those models specifically for predictability.

For a Pro+ subscriber’s monthly premium-request bucket, the Opus 4.6 ceiling falls by 9x. @rayrite ran the math in the community thread: “Opus 4.6 will soon cost 27x after June 1. So you’ll get roughly about 140 or so requests a month.” MindStudio’s worked example puts the same cliff in a cost-per-prompt frame for typical agentic-coding loops.

What’s still murky

• Why the cliff between annual and monthly? GitHub hasn’t published the math behind the spread. The likely read is locked-in revenue meeting rising inference cost, but there’s no posted reasoning.
• Will the new prices freeze in place? The current docs caveat reads, “Model multipliers and costs are subject to change.” Annual subscribers don’t get a locked rate even after this jump.
• What happens to in-flight Pro+ users? GitHub already paused new individual signups on April 20 and pulled Opus from the cheaper Pro tier. Together that means anyone wanting Opus 4.6 access through Copilot now has to pay $39/month and burn 27 premium requests per call.
• Is there a clean opt-out? The docs say a downgrade to monthly avoids the new table, but they don’t spell out whether unused months on the current annual plan are forfeited or pro-rated.

The community thread ran hot. @snakex64 named the motive plainly: “Literally why they’re increasing the annual plan’s premium usage request… so they stop loosing money.” Sentiment that direct doesn’t tend to land in the official changelog, which is part of why the multiplier table arrived as a docs page rather than a blog post.

What this means for you

If you’re an annual Copilot Pro or Pro+ subscriber and you actually use Opus or Sonnet 4.6, do the math before May 31. Multiply your typical monthly request count by the new multiplier, then compare against your premium bucket. Two viable paths: switch to monthly billing, which moves you to GitHub AI Credits priced per token, or cap your usage to the models still on a 1x or 3x multiplier (Sonnet 4, GPT-5.1, Gemini 2.5 Pro). The default of doing nothing means absorbing a 5x to 9x price increase on the model you currently rely on.

For teams whose orgs reimburse individual annual Copilot subscriptions, rebudget now. Cursor’s $20 plan and Anthropic’s $20 Claude Code Pro are the obvious comparisons. Both still get Opus access without a 27x penalty, and neither punishes annual prepayment.