The news
Anthropic released Claude Opus 5 on July 24, 2026, pricing it at $5 per million input tokens and $25 per million output tokens and pitching it as close to the performance of its flagship Claude Fable 5 at half the price.
The price was the same as Opus 4.8's, Anthropic said. Fable 5 cost $10 per million input tokens and $50 per million output tokens, according to Anthropic's pricing page and Trending Topics, double Opus 5's rates on both sides. Anthropic said Opus 5 was available the same day on all of its platforms, including the Claude API under the identifier claude-opus-5. Gizmodo described the release as positioning Opus 5 as a cost-efficient version of Fable 5.
Anthropic built its case on cost per task, not just per-token prices. It said that on CursorBench 3.2, run at maximum effort, Opus 5 scored within 0.5% of Fable 5's peak at half the cost per task. On OSWorld 2.0, which Anthropic described as a computer-use benchmark, the company said Opus 5 beat Fable 5's best result at just over a third of the cost. It also said Opus 5 more than doubled Opus 4.8's score on Frontier-Bench v0.1 at a lower cost per task.
Opus 5 was also offered in a Fast mode that Anthropic said runs about 2.5 times the default speed at twice the base price. On safety, the company said Opus 5 does not advance the frontier in risky dual-use capabilities and scored as its most aligned model to date in an automated behavioral audit. It remains behind Mythos 5 on cybersecurity tasks, Anthropic said; Gizmodo described Mythos 5 as Anthropic's top model, available only to a select group of organizations and government agencies.
Outside testing added nuance. Artificial Analysis, a benchmarking firm that said it evaluated Opus 5 on Anthropic's behalf before release, scored it 61 on its Intelligence Index, ahead of Fable 5 at 60 and GPT-5.6 Sol at 59, according to Trending Topics. The same report said Opus 5's accuracy on the AA-Omniscience test rose seven points over Opus 4.8, but because it answered more often when uncertain, its hallucination rate climbed 14 points to 50%.
The numbers
- Opus 5 API price, input
- $5 per million tokens
- Opus 5 API price, output
- $25 per million tokens
- Fable 5 API price (input/output)
- $10 / $50 per million tokens
- CursorBench 3.2 gap to Fable 5's peak (Anthropic)
- Within 0.5%, at half the cost per task
- Fast mode (Anthropic)
- About 2.5x speed at 2x price
- Hallucination rate on AA-Omniscience (via Trending Topics)
- 50%, up 14 points vs Opus 4.8
Why CEOs should care
For technology and procurement leaders, Opus 5 turns model selection into a routing decision. If Anthropic's cost-per-task figures hold on your own workloads, coding agents and computer-use automations running on Fable 5 are candidates to move down a tier. Run the same evaluation set on both models with your own prompts, measure cost per completed task rather than per token, and keep the flagship only where a quality gap shows up in your results.
CFOs should treat list prices as a starting point. Anthropic's headline coding comparison was measured at maximum effort, and Fast mode doubles the rate, so real bills depend on the settings teams choose. Ask engineering for tokens consumed per task at the effort level you plan to run, then budget by workload.
The Artificial Analysis result is the warning for risk owners and CISOs: a model that answers more often when unsure can produce confident errors. Teams using Opus 5 for research, legal or financial summaries should keep human review and citation checks, and test how the model handles uncertainty before making it a default.
The bigger picture
Opus 5 showed Anthropic competing with its own flagship: a model at half Fable 5's per-token price that, by the company's measures, came close on coding and beat it on computer use. Trending Topics noted that Anthropic sold the release on price and performance rather than as a jump in frontier capability. That narrows the premium a top model can justify and gives buyers more room to negotiate on volume.
What happened next
On September 22, 2026, Anthropic released Claude Opus 5.5 at $4 per million input tokens and $20 per million output tokens. It said the new model cut costs by 40% on typical workloads compared with Opus 5 and generated output more than 30% faster, and that it was raising five-hour usage limits on its Pro, Max, Team and seat-based Enterprise plans.
Anthropic's pricing page now lists Opus 5 and Fable 5 among its legacy models, with Fable 5.1 at $10 and $50 per million tokens as the current top tier. Teams that standardized on Opus 5 in July should rerun their cost-per-task comparisons against Opus 5.5 and watch Anthropic's model deprecation notices for the older version.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made with our newsroom’s technology tools, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is not published automatically; it is held for the editor, who decides whether it is fixed, published or dropped.
- Archive story
- It was written after the event. Its dates were checked against the date of the event, including the dates of any later developments it reports.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story
