Anthropic just did something unusual: it shipped a model nearly as powerful as its flagship, at half the price. On July 24, 2026 it released Claude Opus 5, landing within a hair of Claude Fable 5’s capability — at $5 per million tokens instead of $10. If you run a company and wonder which one fits (or whether this changes your AI costs), here’s the plain-English comparison.
The 60-second summary
- Opus 5 vs Fable 5: Fable 5 is still Anthropic’s most capable released model, but Opus 5 comes within less than 1% on several benchmarks — for half the price.
- Pricing: Opus 5 costs $5 per million input tokens and $25 output. Fable 5 costs $10/$50. Opus 5 does almost the same work at roughly half the cost per task.
- Headline feature: a new 5-level effort control — you tell it how hard to “think” based on the task, and pay accordingly.
- For a business: the flagship is no longer the only serious option. For 90% of use cases, Opus 5 is the smart call on cost.
What each one is, no fluff
Claude Fable 5 was, until now, the most powerful model Anthropic had released publicly: a “Mythos-class” model with reinforced safety, built for the hardest work. Claude Opus 5 is the newest member of the Opus line — the “luxury-but-practical” tier — arriving with the strongest benchmark profile Anthropic has reported for an Opus, with thinking on by default.
The analogy: Fable 5 is the top-tier sports car; Opus 5 is the executive sedan that hits almost the same top speed, burns half the fuel, and works for daily driving. For the racetrack, take the sports car; for everything else, the sedan leaves you just as well-positioned.
The numbers, side by side
| Item | Claude Opus 5 | Claude Fable 5 |
|---|---|---|
| Launch | July 24, 2026 | June 9, 2026 |
| Input price (1M tokens) | $5 | $10 |
| Output price (1M tokens) | $25 | $50 |
| Context window | 1M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens |
| Effort control | 5 levels (new) | Single path |
| Fast mode | Yes ($10/$50, ~2.5× faster) | — |
| Positioning | Near-Fable 5 at half the price | Anthropic’s most capable |
A “token” is roughly ¾ of a word, so one million tokens is about 750,000 words. Both models can read complete contract sets, entire codebases, or a year of customer conversations in one pass.
What the benchmarks say (and what they actually mean)
- Agentic coding (CursorBench): Opus 5 at max effort landed within 0.5% of Fable 5’s best score — at half the cost per task.
- Computer use (OSWorld): Opus 5 beat Fable 5’s best result, at about a third of the cost.
- Against OpenAI’s GPT-5.6 Sol: Opus 5 leads on the evals with public numbers — 43.3% vs 34.4% on frontier agentic coding, and a wide margin on abstract reasoning.
The honest read: at this level, the gap between Opus 5 and Fable 5 is so small it’s invisible for almost any business use. The question is no longer “which is smarter?” but “why would I pay double for a difference I won’t notice?”
The new dial that actually saves money: “effort”
Opus 5’s most practical feature is the 5-level effort control. In plain terms: you tell the model how hard to work based on the task. Low = fast and cheap for routine work (classifying emails, answering FAQs). High = think longer on hard problems (analyzing a contract, planning a migration). You used to pay for the “full brain” on everything; now you pay for what the task actually needs. At volume, that means noticeably lower bills.
What this costs your business
Let’s ground it: an AI chat on your website handling 1,000 conversations a month burns ~10-15 million tokens. Approximate cost:
| Model | Approx. monthly cost | When to use it |
|---|---|---|
| Opus 5 (low effort) | $60–$150 | Customer chat, classification, FAQ |
| Opus 5 (high effort) | $120–$300 | Analysis, task-running agents |
| Fable 5 | $160–$400 | Only the genuinely heavy or strict-compliance work |
The takeaway almost nobody says out loud: your business chat doesn’t need the most expensive model. Opus 5 at low or medium effort handles 90% of cases at a fraction of the cost — you scale up only when the task demands it.
What this race means for your company
- The price of capable AI dropped again. A year ago, “nearly the best model in the world” cost double. Frontier-AI automation is now an operating expense, not a luxury investment.
- Brand matters less than configuration. The winning stack combines the right effort level per task — not “the most expensive model for everything.”
- Build model-agnostic. If your AI system is built right, switching from Fable 5 to Opus 5 (or whatever ships next month) should be a config change, not a rebuild — exactly how we build in custom software development.
- Don’t wait for the next release. This race won’t pause; every month your competitor automates and you don’t, you lose margin.
Finding this analysis useful?
We publish guides like this whenever something big happens in AI and business technology. Leave your email and we'll let you know — no spam, promise.
How to choose: the one-week pilot
- Define 10-20 real tasks from YOUR operation (quoting, customer replies, document summaries).
- Run them through Opus 5 (medium effort) and Fable 5.
- Compare four things: accuracy, annoying refusals, latency and real cost per task.
- Decide on that evidence. In our experience building AI systems for companies, Opus 5 at medium effort wins most of the time.
FAQ
Does Opus 5 replace Fable 5?
It complements it, not replaces it. Fable 5 is still Anthropic’s most capable model for the hardest work; Opus 5 is the cost-smart choice for nearly everything else.
Which is “smarter”?
Fable 5, by a razor-thin margin on the toughest evals. But Opus 5 lands within 1% for half the price — for a business, that gap rarely shows.
Does this change what I pay in ChatGPT or Claude apps?
Subscription apps update on their own. These prices apply when a company integrates the model via API into its systems (like a chat on your site or an internal agent).
Should I wait for Opus 5.1 or the next one?
No. Ship with what exists today, on a system that lets you swap models easily. Waiting has a cost; migrating shouldn’t cost anything if it’s built right.
Sources
- Anthropic — What’s new in Claude Opus 5
- MarkTechPost — Claude Opus 5 launch
- Quartz — Opus 5 at half the price of Fable 5
Want to know which model and effort level fits your operation — and what it would cost per month? At Azterion we implement AI with the right configuration for each task. Book a free 45-minute call and we’ll run the numbers on your business.
