Claude Opus 5
Claude Opus 5 is Anthropic's generally available Opus model released on July 24, 2026, with the API model ID claude-opus-5 and a focus on complex agentic coding and enterprise work.
Opus 5 replaces Opus 4.8 as Anthropic's current Opus recommendation while keeping the same $5 input and $25 output price per million tokens. Anthropic positions it near Fable 5 capability at half Fable's base token price, making it a practical high-end choice for teams that do not need Fable or restricted Mythos access.
Anthropic's announcement and Claude Platform release notes confirm the July 24, 2026 launch, model ID, platform availability, 1M-token context window, 128K maximum output, adaptive thinking, and pricing. Anthropic's performance comparisons are vendor-reported. A high-signal Reddit launch discussion shows reader demand around Opus 5 versus Fable 5, usage limits, migration from Opus 4.8, and real coding behavior, but it is not used for model specifications.
- Complex multi-file coding features, refactors, and difficult debugging.
- Long-horizon agentic workflows that need sustained planning and tool use.
- High-precision code review and bug finding.
- Document, spreadsheet, presentation, and enterprise knowledge work.
- Vision-assisted analysis and frontend replication with iterative tools.
- Migration evaluation from Claude Opus 4.8.
Claude Opus 5 is available across Anthropic's paid Claude plans and developer platforms. The official model overview lists the dateless pinned model ID claude-opus-5, text and image input, text output, a 1M-token context window, 128K maximum output, and availability through the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry.
- Release date: July 24, 2026.
- Base API pricing: $5 per million input tokens and $25 per million output tokens.
- Adaptive thinking is on by default, with effort levels from low through max.
- Fast mode is a separate research preview on the Claude API at premium pricing.
Opus 5 is the advanced everyday option for difficult coding, analysis, and enterprise workflows. Fable 5 remains Anthropic's highest-capability generally available tier at twice the base token price, while Mythos 5 is limited to approved Project Glasswing customers. Sonnet 5 is the lower-cost default for workloads that prioritize speed and production scale.
- Choose Opus 5 when task success on hard agentic work matters more than lowest token cost.
- Choose Fable 5 only when evaluations justify the higher price for the most demanding long-running work.
- Treat Mythos 5 as a restricted-access option, not a normal public upgrade path.
- Evaluate Sonnet 5 first for routine production work where speed and cost dominate.
Opus 5 is designed to work with existing Opus 4.8 prompts, but it changes important runtime behavior. Thinking is enabled by default, disabling thinking is allowed only at high effort or below, and Anthropic recommends using effort as the main cost-and-depth control. Re-run prompt, tool, refusal, and completion evaluations before changing production routing.
- The base token price is unchanged from Opus 4.8.
- The minimum prompt-cacheable length drops to 512 tokens.
- Mid-conversation tool changes and automatic fallbacks are beta features with dated headers.
- Fast mode support is limited to the first-party Claude API and is not a different model.
Launch discussion centers on whether Opus 5 delivers enough real coding and agent improvement to replace Opus 4.8, why Fable 5 still costs more, how paid-plan usage limits behave, and whether lower effort levels preserve quality. Those are useful evaluation questions, but early anecdotes should not be treated as benchmark proof.
The previous Opus release and primary migration baseline.
Anthropic's higher-capability generally available tier at twice Opus 5's base token price.
Restricted-access model for approved Project Glasswing customers.
The faster, lower-cost Claude 5 option for production workloads.
Source confidence
Anthropic
Anthropic
Anthropic
Anthropic
Anthropic
Anthropic
Claude Opus 5 FAQ
Common questions about Claude Opus 5.
Is Claude Opus 5 officially released?+
Yes. Anthropic released Claude Opus 5 on July 24, 2026. It is available on paid Claude plans, Claude Code, the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry.
What is the Claude Opus 5 API model ID?+
The Claude API model ID is claude-opus-5. Anthropic documents it as a dateless pinned snapshot rather than an evergreen alias.
How much does Claude Opus 5 cost?+
Claude Opus 5 costs $5 per million input tokens and $25 per million output tokens on the standard Claude API. Prompt caching, batch processing, and Fast mode use separate rates.
Should I choose Claude Opus 5 or Claude Fable 5?+
Start with Opus 5 for difficult everyday coding and enterprise work, then move to Fable 5 only when your own evaluations justify its higher capability and twice-the-base-token price. Anthropic positions Fable 5 as the highest-capability public tier and Opus 5 as the more efficient advanced tier.
Can I migrate directly from Claude Opus 4.8 to Opus 5?+
Yes, but re-test the workflow before changing production traffic. Opus 5 works with existing Opus 4.8 prompts, while thinking defaults, effort controls, tool behavior, and some beta headers can change runtime results.