Claude Opus 4.8 is Anthropic's latest and most capable Opus model. $5/$25 per MTok, 1M context, vision support, and improved reasoning over Opus 4.7. Full pricing and API access details.
Claude Opus 4.8 is the latest update to Anthropic's Opus model family, released in August 2026. It builds on Opus 4.7 with improved reasoning, better instruction following, and more reliable tool use. The model features a 1,000,000-token context window and supports both text and image inputs.
Opus 4.8 is designed for the most demanding tasks — complex multi-step reasoning, long-document analysis, advanced coding, and research workflows. It sits above Claude Sonnet 5 in capability, with a price point that reflects its premium positioning.
Claude Opus 4.8 costs $5 per million input tokens and $25 per million output tokens at standard rates. This is unchanged from Opus 4.7 — the improvement is in capability, not price. Prompt caching is supported, with cache reads at $0.50/MTok and cache writes at $6.25/MTok.
| Model | Input (cache miss) | Input (cache hit) | Output | Context |
|---|---|---|---|---|
| Claude Opus 4.8 (claude-opus-4-8) | $5.00 | $0.50 | $25.00 | 1,000,000 |
Source: Anthropic official pricing as of August 2026. Batch API available at 50% of Standard rates.
Opus 4.8 offers improved reasoning and instruction following over Opus 4.7, at the same $5/$25 pricing. For most workloads, the upgrade is a drop-in replacement — change the model ID from claude-opus-4-7 to claude-opus-4-8 and the API format remains identical.
Compared to Claude Sonnet 5 ($2/$10 with promo pricing), Opus 4.8 provides stronger reasoning on complex tasks but costs 2.5x more per token. For straightforward coding, analysis, and chat tasks, Sonnet 5 is typically sufficient. Opus 4.8 shines on tasks that require multi-step planning, long-context synthesis, or the highest accuracy.
At $5/$25 per MTok, Opus 4.8 is priced below GPT-6 Astra ($10/$50) and GPT-5.6 Sol ($5/$30), and above Claude Sonnet 5 ($2/$10) and GPT-5.4 ($2/$8). Prompt caching significantly reduces costs for agentic workloads — at 90% cache hit rate, effective input cost drops to about $0.95/MTok.
For developers evaluating between Opus 4.8 and competing models, the key differentiator is Opus 4.8's combination of 1M context, vision support, and strong instruction following. GPT-5.6 Sol offers similar capability at a comparable price point, while GPT-5.4 provides a lower-cost alternative for tasks that don't require Opus-level reasoning.
Claude Opus 4.8 is available through the Anthropic API and on AWS Bedrock and Google Vertex AI. It uses the standard messages API format, so existing Claude integrations can upgrade by changing the model ID to claude-opus-4-8.
APItokendeal provides OpenAI-compatible access to Claude Opus 4.8 and other Anthropic models through a single API key. Teams can use token packs with prepaid credits, getting 77% savings compared to direct API billing at pay-as-you-go rates.
APItokendeal provides OpenAI-compatible API access to Claude Opus 4.8, GPT, DeepSeek, Gemini, Kimi, and more. One API key, one dashboard, prepaid token packs with no surprise bills.