Claude Opus 4.8
FrontierLast verified: September 25, 2026· Anthropic Claude Opus 4.8 pricing
Previous-generation Opus, still served. Opus 5.5 is cheaper per token ($4 / $20). Adaptive thinking, 1M context.
- Input / 1M
- $5.00
- Output / 1M
- $25.00
- Context
- 1M tokens
- Provider
- Anthropic
Pricing verified September 25, 2026. Prices change frequently. Always confirm against the provider’s official pricing page before relying on these figures for budgeting. Official pricing →
When to choose Claude Opus 4.8
The default frontier choice for agentic coding and knowledge work. What makes it a default rather than a fallback is the pricing shape: 1M tokens of context at standard rates with no long-context surcharge, so a large prompt costs what the arithmetic says and nothing more. At $5 in and $25 out it is a mid-priced frontier row rather than the top one, which puts frontier-tier capability within reach of work that runs all day rather than work that runs in a demo. The reply is capped at 128K tokens.
Claude Opus 4.8 sits close enough to Claude Fable 5 on price that the two are usually decided together. That comparison is written once, on the models table, rather than argued twice from opposite sides.
What Claude Opus 4.8 costs in practice
A 50M-input, 10M-output month lands at $500. The comparison that bites is not inside the Claude range but across it: GPT-4.1 carries a window of the same size and does that month for $180, so the real question is whether the capability difference is worth 2.8x on your own routes. Filling the whole 1M window costs $5 of input, and because there is no long-context premium that figure stays linear as prompts grow, which makes large-context work something you can budget rather than a step change discovered at invoice time. The reply is capped at 128K tokens, so a genuinely long generation is a batching problem rather than a single call.
| Scenario | Input / mo | Output / mo | Est. cost / mo |
|---|---|---|---|
| Prototype | 2M | 0.5M | $23 |
| Growing product | 50M | 10M | $500 |
| At scale | 500M | 100M | $5,000 |
Plug in your own numbers with the cost calculator.
Where Claude Opus 4.8 fits in a routing setup
Routing around this row runs in both directions. Downward, Sonnet 4.6 does the reference month for $300 against $500 and absorbs a large share of production traffic that never needed the frontier tier. Upward, Fable 5 exists for the runs Opus cannot finish. Sideways, GPT-4.1 offers a window of the same size for $180, so work that is context-bound rather than capability-bound may not need this row at all. Which of the three moves applies is a per-route question, and evals are how it gets answered.
Cheaper alternatives to Claude Opus 4.8
The models immediately below Claude Opus 4.8 on blended cost, nearest first. A swap only pays if quality holds on the routes you move — see the LLM cost optimization playbook for how to prove that before you move them.
Frequently asked questions
How much does Claude Opus 4.8 cost?
Claude Opus 4.8 bills $5.00 per million input tokens and $25.00 per million output tokens — $10.00 per million blended at a three-to-one input:output split. A workload of 50M input and 10M output tokens per month works out at about $500.
What is Claude Opus 4.8's context window?
Claude Opus 4.8 has a 1M-token context window, of which up to 128K tokens can be the reply. One request that fills the whole window costs about $5.00 at the input rate, before the reply is billed.
Where does Claude Opus 4.8 sit on price?
At $10.00 per million blended — three input tokens for every output token — Claude Opus 4.8 is dearer than 16 of the 19 models tracked here and cheaper than 2. Output bills at 5x the input rate, and the nearest cheaper model on this table is Claude Opus 5.5 at $8.00 blended.
Related concepts: context window, model routing, prompt caching, evals.