Skip to content

Claude Opus 4.7 lands in GitHub Copilot at a 7.5x premium request rate

Claude Opus 4.7 lands in GitHub Copilot at a 7.5x premium request rate

Claude Opus 4.7 is rolling out inside GitHub Copilot, and it arrives with a significant catch: each request now burns 7.5 premium request units instead of the 3x rate that Opus 4.6 carried. The new model is available to Copilot Pro+, Business, and Enterprise users, and it will gradually replace both Opus 4.5 and Opus 4.6 in the Pro+ model picker over the coming weeks.

Quick answer: Opus 4.7 is live in Copilot Pro+, Business, and Enterprise at a 7.5x premium request multiplier (promotional through April 30, 2026). Opus 4.5 and 4.6 are being phased out of the Pro+ picker. Pro, Free, and Student plans do not get Opus 4.7.


Where Opus 4.7 is available

The model shows up in the standard Copilot surfaces, though rollout is gradual and some users may not see it immediately. Copilot Business and Enterprise administrators need to flip the Claude Opus 4.7 policy on in Copilot settings before it appears for their users.

SurfaceOpus 4.7 available
Visual Studio CodeYes
Visual StudioYes
Copilot CLIYes
GitHub Copilot Coding AgentYes
github.comYes
GitHub Mobile (iOS and Android)Yes
JetBrains IDEsYes
XcodeYes
EclipseYes

If Opus 4.7 is missing from the Copilot CLI model list, updating to a newer build is the fix. The 1.0.29 release did not include the model in /models during the initial rollout window.


The 7.5x multiplier, explained

Every Copilot paid plan has a monthly allowance of premium requests. Each model consumes a set number of those requests per turn, defined by its multiplier. Opus 4.6 sat at 3x. Opus 4.7 launches at 7.5x, meaning a single chat turn now draws 7.5 premium requests from your monthly quota instead of 3.

On a Pro+ plan with 1,500 premium requests, that works out to about 200 Opus 4.7 turns per month before the allowance runs out, compared to 500 turns at the old Opus 4.6 rate. The 7.5x figure is promotional pricing that GitHub says will stay in place until April 30, 2026. The post‑promotion multiplier has not been announced.

⚠️
The Anthropic API price for Opus 4.7 is identical to Opus 4.6: $5 per million input tokens and $25 per million output tokens. The higher Copilot multiplier reflects Copilot's per‑request pricing model, not a change in underlying token cost.

Multiplier comparison

ModelPaid plan multiplier
Claude Opus 4.77.5 (promotional until Apr 30)
Claude Opus 4.6 (fast mode, preview)30
Claude Opus 4.63
Claude Opus 4.53
Claude Sonnet 4.61
Claude Sonnet 4.51
Claude Haiku 4.50.33
GPT-5.41
GPT-5.4 mini0.33
GPT-5.3-Codex1
Gemini 3.1 Pro1
Grok Code Fast 10.25

Which plans get Opus 4.7

Opus 4.7 is restricted to the top tiers. Copilot Pro, Copilot Free, and the recently introduced Copilot Student plan do not have access to it in the model picker. Students lost self-selection access to Claude Opus and Sonnet models, along with GPT‑5.4, back on March 12, 2026, when GitHub moved verified students onto the separate Copilot Student plan. Those users can still reach Anthropic and OpenAI models indirectly through Auto mode, but cannot pick Opus 4.7 manually.

PlanOpus 4.7 in picker
Copilot FreeNo
Copilot StudentNo
Copilot ProNo
Copilot Pro+Yes
Copilot BusinessYes (admin must enable)
Copilot EnterpriseYes (admin must enable)

What changes in the model itself

Opus 4.7 is positioned by Anthropic as a step up from Opus 4.6 on multi-step tasks, long-horizon reasoning, and agentic coding work where the model has to chain tools together across a longer session. It handles ambiguous prompts with less hand-holding and is described as better at bug-finding and code review.

Two behavior changes stand out for anyone migrating prompts or harnesses from 4.6:

  • Adaptive thinking replaces a fixed thinking budget. Extended Thinking with a manual token budget is not supported. The model decides when to reason more and when to answer directly, which Anthropic says cuts down on the overthinking behavior that 4.6 was prone to.
  • Fewer tool calls and fewer subagents by default. Opus 4.7 leans harder on internal reasoning. If a workflow benefited from aggressive file reads, parallel subagents, or frequent tool invocation under 4.6, the prompt now needs to spell that out explicitly.

Response length is also calibrated more tightly to task complexity, so simple lookups get shorter answers than 4.6 would have produced and open-ended analysis can run longer.


Reasoning effort and context window in Copilot

Inside Copilot specifically, Opus 4.7 is currently locked to medium reasoning effort. The model picker does not expose a reasoning selector for it, and the context window is capped at 200K tokens. That is a notable downgrade for anyone who had been using Opus 4.6 at high effort with the 1M context variant for large codebase work.

Anthropic's own Claude Code product exposes additional effort levels (low, medium, high, xhigh, max), with xhigh as the new default. Those settings are not mirrored inside GitHub Copilot at launch.


Opus 4.5 and 4.6 sunset timing

GitHub has said Opus 4.7 will replace both Opus 4.5 and Opus 4.6 in the Copilot Pro+ picker over the coming weeks. Neither model has a confirmed retirement date published on GitHub's model retirement list yet, so the exact sunset dates for 4.5 and 4.6 inside Copilot remain unannounced. The retirement history page currently lists earlier models like Opus 4.1 (retired February 17, 2026) as the most recent Anthropic removals.

Pro+ users who relied on Opus 4.6 at 3x for everyday work are facing a real cost shift. A workflow that used to spend 3 premium requests per turn on Opus 4.6 now spends 7.5 on Opus 4.7, with no guaranteed drop after the promotional window ends on April 30.


Alternatives inside Copilot

If the 7.5x multiplier is too steep for your volume of work, the practical alternatives inside Copilot itself are the Sonnet and GPT tiers, which remain at 1x.

Use caseLower-cost option in CopilotMultiplier
General coding and chatClaude Sonnet 4.6 or GPT-5.41
Codex-style code generationGPT-5.3-Codex or GPT-5.2-Codex1
Fast, cheap turnsClaude Haiku 4.5 or GPT-5.4 mini0.33
Fastest flat-rate optionGrok Code Fast 10.25

Sonnet 4.6 covers most day-to-day coding tasks at a third of Opus 4.6's cost and roughly one-seventh of Opus 4.7's current cost. For workflows that genuinely benefit from Opus-class reasoning, there is no direct substitute inside Copilot right now.


How to know Opus 4.7 is actually selected

The model picker in VS Code, Visual Studio, JetBrains, Xcode, Eclipse, Copilot CLI, github.com, and GitHub Mobile will show Claude Opus 4.7 as a selectable entry once the rollout reaches your account and (for Business/Enterprise) once an admin has enabled the policy. If it is missing from Copilot CLI, make sure you are on a build newer than 1.0.29 and run /models again after updating.

Premium request consumption is visible in your GitHub billing and usage dashboard. A turn on Opus 4.7 will deduct 7.5 units; if you see a 3-unit deduction instead, the request was routed to Opus 4.6 rather than 4.7.

For most Pro+ subscribers, the real decision over the next few weeks is not whether Opus 4.7 is better than Opus 4.6, but whether it is 2.5 times better, because that is the ratio the Copilot multiplier is currently asking you to pay.