Anthropic released Claude Opus 5 on July 24, and the headline detail is the price: $5 per million input tokens and $25 per million output tokens (around ₱310 and ₱1,550). That is unchanged from Opus 4.8 and half of what the company charges for Fable 5, its most powerful model. Anthropic describes Opus 5 as "a thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price," and says it is available today across all platforms.
A token is roughly three-quarters of an English word, so a million tokens is about 750,000 words — the pricing above is what it costs to feed in or get back that much text through the API.
What is new in Claude Opus 5
Opus 5 ships with a 1 million-token context window, meaning it can hold about 750,000 words of documents, code, or conversation in a single session. Maximum output is 128,000 tokens. Its knowledge cutoff is May 2026, the most recent of any Claude model, so it needs less prompting about events from earlier this year.
Anthropic points to gains in software engineering, scientific work such as organic chemistry and protein analysis, visual generation, and agentic reasoning — where the model plans a multi-step task and checks its own work as it goes.
How Claude Opus 5 scores
The benchmark results Anthropic published are the clearest argument for the model:
- Frontier-Bench v0.1 — Opus 5 more than doubles Opus 4.8's score and comes out ahead of every other model tested.
- CursorBench 3.2 — lands within 0.5% of Fable 5's peak score at half the cost per task.
- ARC-AGI 3 — scores three times as high as the next-best model on novel problem-solving.
- OSWorld 2.0 — surpasses Fable 5's best result at roughly a third of the cost.
These are the developer's own published numbers. Independent evaluators have started weighing in: Quartz reported that Opus 5 took the top spot on Artificial Analysis's Intelligence Index at 61 and its Agentic Index at 55.3.
Price against the rest of the Claude line
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|
| Claude Opus 5 | $5 (around ₱310) | $25 (around ₱1,550) |
| Claude Opus 4.8 | $5 (around ₱310) | $25 (around ₱1,550) |
| Claude Fable 5 | $10 (around ₱620) | $50 (around ₱3,100) |
Peso figures use a rate of about ₱61.86 to the dollar and are rounded.
The safety numbers Anthropic published
Anthropic reported that on its automated behavioural audit, "Opus 5 scores 2.3 on overall misaligned behavior, the lowest of our recent models." It also states plainly that Opus 5 still trails Mythos 5 on cybersecurity exploitation tasks. Publishing a capability the model is behind on is unusual enough to be worth noting, though these remain self-reported results from the company that built it.
What it changes for users in the Philippines
There is no Philippine-specific pricing here — Anthropic bills the API in US dollars, so local developers pay the converted rate plus whatever their card issuer adds. For everyday users, the more relevant route is the bundle: Globe packages Claude alongside ChatGPT, Gemini and Grok in a single app from ₱49, which is far cheaper than a standalone subscription for someone who just wants to try a frontier model.
For small development shops here, the cost line is the real story. If a task that previously needed the most expensive tier now runs acceptably on Opus 5, the bill for that workload roughly halves — and Anthropic's own OSWorld and CursorBench figures are built to make exactly that argument. Whether it holds for your workload is something to test rather than assume.
Access remains uneven by region. Anthropic cut off Chinese access to Claude earlier this year, a reminder that availability of these models is a policy decision as much as a technical one.
