TL;DR: Ghost has upgraded its default model from Claude Sonnet 4.6 to Claude Opus 4.8 — Anthropic's most capable reasoning model. The complex prompting Ghost uses to analyse top-performing LinkedIn posts and extract pattern-level insights now runs against a model that handles longer context, reads voice and tone with more nuance, and writes back in a way that sounds like a human who has actually studied your industry rather than an AI guessing at it.

Ghost is a LinkedIn GTM platform that connects content creation to intent-powered outbound. The content engine — the part of Ghost that learns your brand voice, studies what is working in your niche right now, and generates posts in your voice — has run on Claude since day one. Today it's running on Claude Opus 4.8, and the difference in output quality is the biggest single jump we've shipped this year.

What changed

Up until this week, Ghost's content engine ran on a mix of Claude Sonnet 4.6 for fast generation and Claude Opus 4.6 for the heavy reasoning steps — voice modelling, content analysis, and the post-rewriting passes. As of this release, the default reasoning model across the platform is Claude Opus 4.8, Anthropic's flagship model released earlier this quarter.

For users on a Ghost plan, this happened invisibly. Your existing content settings, brand voice profile, and posting cadence carried over unchanged. What you'll notice is that the suggestions feel sharper, the post drafts need less editing, and the insights Ghost surfaces about why a particular post performed well are more specific and more useful.

Why this upgrade matters specifically for Ghost

Most AI content tools use a foundation model to write a post from a prompt. Ghost does something more involved — and that complexity is why the model upgrade actually moves the needle for our users.

Every time Ghost generates a post for you, it runs a chain of prompts that includes:

  • Your brand voice profile — the patterns extracted from your existing LinkedIn posts, comments and articles
  • A live pull of top-performing posts in your niche — what your competitors and peers are getting traction with right now
  • Your content pillars — the strategic themes you've defined for your audience
  • The engagement data from your own past posts — what hooks, formats and topics actually moved the needle for you specifically
  • The commercial context — your ICP, your product, the buyer concerns you're addressing

That's a lot of context to hold simultaneously, and a lot of reasoning to do in a single inference. Sonnet 4.6 handled it well. Opus 4.8 handles it noticeably better.

Founder's Take: The thing I underestimated about model upgrades early on is that they don't just make the writing slightly better — they make complex chained prompts work where they previously broke down. The voice modelling step in Ghost is where I see the biggest difference. Opus 4.8 can hold your brand voice across a 30-post analysis without drifting toward "AI-default" tone in the way earlier models did.

What Claude Opus 4.8 actually brings to the table

Anthropic's headline numbers for Opus 4.8 cover the obvious benchmarks — reasoning, code, mathematical performance. For Ghost specifically, three capabilities matter most:

Longer effective context, applied properly

Opus 4.8 has a 500K-token context window, but the more important number is the effective context — how well the model uses information at the start of a prompt by the time it's writing the output. Earlier models would receive 80 of your past LinkedIn posts as context, then write a new post that ignored half of them. Opus 4.8 references and synthesises across the full set, which means your generated posts actually sound like they were written by someone who has read everything you've ever published — because the model has, and is now able to use that.

Voice and tone fidelity

The hardest thing for any AI model is sustaining a specific, distinctive writing voice across multiple outputs. Generic AI tone is what happens when the model loses track of the constraints and defaults to its training average. Opus 4.8's voice fidelity is a meaningful step up — in our internal testing, customers couldn't reliably tell Ghost-generated drafts from their own first-pass writing on 7 of 10 tested voices.

Pattern recognition across competitor content

One of the most valuable things Ghost does is look at the top 50–100 posts in your niche from the last 30 days and extract the patterns that are driving engagement right now — not generic "use a hook" advice, but specific patterns like "in the early-stage SaaS founder niche this month, posts that open with a £-figure revenue claim and end with a contrarian take on outbound are converting at 3x baseline." Opus 4.8 surfaces these niche-specific patterns with much more precision because it can hold all 100 posts in working memory and reason across them rather than sampling.

What you'll see change in Ghost

The upgrade affects three places in the product:

The post generator

Drafts come out closer to ready-to-publish. The "first edit" pass that most users did before posting is shorter, and in some cases unnecessary. The model is better at picking up specific verbal tics from your brand voice — em-dashes, sentence fragments, particular phrasings you reuse — and applying them naturally rather than mechanically.

The content insights panel

When Ghost analyses why a competitor's post performed well, the explanation is now more specific. Instead of "this post used a strong hook," you'll get "this post inverted the expected framing on day three of the week, when audience attention is shifting from operational content to strategic content, and the contrarian framing benefits from that timing window." Those are the kinds of insights you can actually apply.

The Agentic Campaign Creator

The AI agent that builds full outbound campaigns end-to-end — finding the contacts, writing the sequence, building the campaign — uses Opus 4.8 for the sequencing reasoning. The personalised first lines and follow-up messages now reference engagement signals more naturally. Less "I noticed you liked this post," more "your point on attribution in last week's comment thread is exactly what I wanted to push back on" — sequencing that sounds like a thoughtful peer rather than a templated bot.

What this is not

To be straight about it: this upgrade does not change what Ghost is. We are still a LinkedIn GTM platform with content + intent + multi-channel outbound + a 600M contact database. The model upgrade improves how well the content engine does what it already did — it doesn't introduce new features.

And we're not claiming Opus 4.8 generates posts indistinguishable from human writing. Strong writers still write better than AI on the highest-stakes posts. What Ghost does is take the 90% of LinkedIn posting that is mechanical, time-consuming, and not worth a founder's own hours, and make it 90% of the way to "as good as you'd write yourself." For most B2B founders running their own GTM, that frees up the four hours a week LinkedIn was eating without losing the pipeline it was generating.

How model routing works inside Ghost

Ghost integrates with OpenRouter, which means under the hood you can switch between Claude Opus 4.8, Claude Sonnet 4.6, GPT-5, Gemini 2.5 Pro, and Grok 4. Opus 4.8 is the new default for content reasoning, but if you want to test alternative models for specific tasks — say, GPT-5 for technical posts, Gemini for image-heavy content — that's exposed in the model settings.

For 95% of users the right answer is to leave the default on Opus 4.8 and not think about it. The model routing exists for the power users who want to A/B test approaches.

Frequently asked questions

Do I need to change any settings to get the upgrade?

No. Claude Opus 4.8 is the new default for all Ghost users from this week. Your existing brand voice, content pillars, and posting schedule carry over unchanged. The model is selected automatically for each reasoning step.

Will my generated posts cost more or take longer to produce?

No. Ghost's pricing is unchanged — still $99/month (≈£75) for the full platform with everything included. Generation time is comparable to before; Opus 4.8 is marginally slower per token but the chained-prompt structure inside Ghost means total time per post is unchanged.

What about my existing brand voice profile — will it still work?

Your brand voice profile is model-agnostic — it's a structured representation of your writing patterns, not a model-specific prompt. The same profile works against Opus 4.8 and will give you noticeably better fidelity than it did against earlier models.

How does Ghost pull in top-performing LinkedIn content for analysis?

Ghost analyses public LinkedIn content from accounts in your defined niche, scored by engagement metrics (likes, comments, shares, reposts). The model reasons over the post bodies, structures, and engagement patterns to extract what's working right now — surfaced to you as both content suggestions and strategic insights.

Can I roll back to Sonnet 4.6 if I prefer it?

Yes. Model selection is configurable in Settings → AI Models. Opus 4.8 is the default but you can pin any specific generation step to a different model. We expect most users won't need to.

Does this affect intent signal scoring?

The intent scoring system itself runs on a separate signal model — not Claude. The upgrade affects content generation, content analysis, and the Agentic Campaign Creator. Intent scoring continues to use the same first-party signal pipeline.

When will Opus 4.9 be available?

When Anthropic ships it. Ghost runs on OpenRouter for model access, which means we typically integrate new flagship models within 48 hours of release. You'll see the model bump in your settings panel automatically.

Start your 7-day Ghost trial to see the Opus 4.8 content engine in action — full platform access, no credit card required.