Claude Opus 4.8
Claude Opus 4.8 is about 4x less likely than Opus 4.7 to let flaws in its own code pass unremarked, at unchanged $5/$25 pricing.
Upgrade on Opus 4.7 in coding, agentic skills, reasoning and knowledge work, with more honest flagging of uncertainty. Shipped with Claude Code dynamic workflows (hundreds of parallel subagents), a claude.ai effort control, system messages inside the Messages API array, and fast mode at 2.5x speed for $10/$50.
- Date
- Thursday, 28 May 2026
- Lab
- Anthropic
- Kind
- model
- Access
- closed API
- Price
- $5/$25 per M tokens in/out, unchanged from Opus 4.7; fast mode $10/$50 (2026-05-28)
Figures
| Measure | Value | Measured by |
|---|---|---|
| Flaws in own code left unremarked | about 4x less likely vs Opus 4.7, per Anthropic evaluations | company |
| Online-Mind2Web (browser agent) | 84% A tester quote puts it above Opus 4.7 and GPT-5.5. | third-party |
Anthropic calls it a 'modest but tangible improvement' and says a higher class of model, plus Mythos-class models for all customers, would follow 'in the coming weeks'; Fable 5 arrived 12 days later with Opus 4.8 as its fallback model. The launch post benchmarks against GPT-5.5 (Terminus-2 harness; GPT-5.5 reports 83.4% on Terminal-Bench 2.1 with Codex CLI) and Gemini 3.5 Flash. Per-benchmark Opus 4.8 scores sit in an image table that was not machine-readable.
Sources
- www.anthropic.com/news/claude-opus-4-8
- claude.com/blog/introducing-dynamic-workflows-in-claude-code
- platform.claude.com/docs/en/release-notes/overview
This record was checked against its sources on 6 October 2026. How we check