Opus 5 vs 4.8 vs 4.7 vs 4.6: comparing revisions priced identically

Claudexia currently serves four Opus revisions: 5, 4.8, 4.7, and 4.6. Model choice usually starts with price, but that lever doesn't exist here, the price is identical across all four. So the choice comes down to behavior and stability, not the bill.

All four revisions side by side

For context: Anthropic's own list price for Opus 5 is $5 per million input tokens and $25 per million output tokens. Through Claudexia all four Opus revisions run at $0.4 and $0.4, regardless of which one you call.

If the price is the same, what's the choice about

If price scaled with the revision number, picking one would be a quality-versus-budget tradeoff. That axis doesn't exist for us, so the decision comes down to three questions: do you need the flagship's speed and new capabilities, do you need the predictability of behavior you already tuned around, or do you need the reproducibility of a frozen version for evals and audits.

When to pick the newest one

Opus 5 shipped two concrete improvements: noticeably faster responses on long agentic runs, and the ability to change its tool set mid-conversation without starting a new session. Plus a built-in fallback to 4.8 if the safety classifier flags a request, so even calling 5 directly gets you that safety net without extra code.

Default to 5 for new projects unless there's a reason to hold onto an older revision.

When stability matters more than novelty

• Production is already tuned to a specific revision. Prompts, few-shot examples, response parsing, all of that gets calibrated to a model's behavior. Swapping revisions without testing is a regression risk you get for free and without warning.
• You have a frozen eval dataset. Pinning fixes the comparison baseline over time. If a pass-rate metric needs to be comparable month over month, the model under the hood can't change.
• You need documented reproducibility. In audits or regulatory processes it sometimes matters to know exactly which model version handled a given request.

For these scenarios, 4.8, 4.7, or 4.6 work exactly as well as before, and carry no price penalty for not being the flagship.

Switching revisions in code

One line changes, model, the rest of the code stays untouched:

from openai import OpenAI

client = OpenAI(api_key="sk_cdx_...", base_url="https://api.claudexia.tech/v1")

for model in ["claude-opus-5", "claude-opus-4.8", "claude-opus-4.7", "claude-opus-4.6"]:
    resp = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": "Draft a migration plan for the service"}],
    )
    print(model, "->", resp.choices[0].message.content[:80])

Running the same case across all four revisions is the most honest way to see the behavioral difference on your own task rather than someone else's description of it.

A practical selection scheme

1. New project with no legacy constraints, take claude-opus-5.
2. Production is already tuned and working, keep the current revision until there's a reason to migrate.
3. You're mid-comparison before a migration, keep production on the older revision and run your eval dataset against the new one in parallel.
4. You need a documented version for audit purposes, pin to a specific id and set up a recurring check on that revision's status with Anthropic.

There's no need to model the price difference between options, there isn't one. Spend per key and per model shows up in the dashboard, and the calculator on the site estimates your bill for your own token volumes.

In short

All four Opus revisions cost the same with us: $0.4 on input and $0.4 on output per million tokens. Opus 5 is faster and can swap tools mid-flight, 4.8 covers the fallback and workhorse role, 4.6 and 4.7 stay in the lineup for teams that need reproducibility. Pick by task, not by bill, the bill doesn't move here.