Claude Opus 5 arrives with near-Fable coding and new self-checking behavior
Anthropic released Claude Opus 5 on July 24, 2026. Based on AlexKim’s migration-guide notes, pricing appears unchanged from Opus 4.8 at $5 per million input tokens and $25 per million output tokens, while programming performance is described as very close to Fable 5, with roughly a 0.5-point gap. The most notable update is not just benchmark performance but a behavior shift: Opus 5 reportedly self-verifies answers without being explicitly told, which could make some long-standing prompting and verification practices less useful.
Confirmed
- AlexKim says Opus 5 keeps Opus 4.8 pricing: $5 per million input tokens and $25 per million output tokens.
- Multiple posts describe it as “Fable-class intelligence”; AlexKim says its coding ability is very close to Fable 5, with about a 0.5-point gap.
- Aidenybai says Opus 5 is the strongest Claude-family model on ReactBench and highlights its frontend coding value.
- Majidmanzarpour, citing Anthropic’s prompt guide, says Opus 5 is best suited for complex agentic programming tasks and code review, and that prompting should be more constrained.
- AlexKim reports that Opus 5 now supports all five effort levels, and that even lower effort settings are performing surprisingly steadily.
- AlexKim also highlights several API migration gotchas: on Claude 4.8, thinking is still enabled by default if no parameter is passed; on Opus 5, disabling thinking while setting xhigh or max effort returns a 400 error; and maxtokens now caps both thinking tokens and output tokens.
- AlexKim says Opus 5 checks its own answers even without explicit instruction. As a result, older prompts such as “double-check before answering” may cause duplicated work, and some harness-level verification passes may no longer justify their cost.
Unconfirmed
- Some posts from heypearlai and a forwarded post by GregCook2011 claimed Opus 5 was priced at roughly half of Opus 4.8. That conflicts with AlexKim’s migration-guide-based pricing details and is not corroborated elsewhere in the provided materials.
- Heypearlai also said Opus 5 became the new default model for Claude Max, but no direct first-party confirmation is included in the posts provided here.
Why it matters
Opus 5 matters because the reported performance gain comes without a confirmed API price increase, narrowing the gap between Anthropic’s regular flagship line and Fable-class capability in coding tasks. More importantly, the model’s self-checking behavior suggests a workflow change for developers: prompt recipes, verification layers, and token budgeting strategies that worked for earlier Claude versions may now be inefficient or even counterproductive.
2026-07-27 ~ 2026-07-28 · 11 related posts
- Episode 1: Anthropic's Messy Releases Put Pressure on Opus 5(2026-07-23, 2 posts)
- Episode 2: Anthropic Releases Claude Opus 5: SOTA Performance at Half the Price(2026-07-25, 128 posts)
- Episode 3: Anthropic Rumored to Release Opus 5 with Fast Mode and Advanced Visuals(2026-07-25, 3 posts)
- Episode 4: Claude Opus 5 Early Tests: Better Efficiency but Overly Proactive(2026-07-25, 29 posts)
- Episode 5: Anthropic Releases Claude Opus 5 with Impressive Benchmark Results(2026-07-25, 3 posts)
- Episode 6: Claude Opus 5 Accused of Benchmark Gaming, Lags Behind in Real Tests(2026-07-25, 2 posts)
- Episode 7: Claude Opus 5 Sets New ARC-AGI-3 Record(2026-07-25, 15 posts)
- Episode 8: Anthropic Internal Docs Reveal Opus 5 Progress(2026-07-25, 2 posts)
- Episode 9: Opus 5 Impressions: Stunning Single-Prompt Generation but Lags Behind Fable in Complex Tasks(2026-07-26, 24 posts)
- Episode 10: Claude Opus 5 arrives with near-Fable coding and new self-checking behavior(2026-07-27, 11 posts)
- Episode 11: Anthropic Opus 5 Leads Benchmarks but Splits Real-World Reviews(2026-07-28, 6 posts)
- Episode 12: Claude Opus Series Accused of Degraded Experience: Laziness and Amnesia Spark Trust Crisis(2026-07-29, 14 posts)
- Episode 13: Anthropic Launches Claude Opus 5 with Top Performance at Half the Cost(2026-07-31, 2 posts)
- Episode 14: Anthropic Faces Developer Backlash Over Declining Model Performance(2026-08-03, 11 posts)
Primary sources
- Claude Opus 5 arrives at half the price and tops Frontier-Bench claims — GregCook2011 · 2026-07-27
- Anthropic ships Claude Opus 5 at half the price of the old Opus 4.8 tier — heypearlai · 2026-07-27
- Anthropic’s Opus 5 tops a ReactBench claim while costing more than 2× less — aidenybai · 2026-07-28
- Claude Opus 5’s migration guide quietly changes years of prompting advice — AlexKim · 2026-07-28
- [source] Anthropic’s Opus 5 now self-checks by default, making old “double-check” prompts redundant — AlexKim · 2026-07-28
- Anthropic says older harness-level verification may now be redundant — AlexKim · 2026-07-28
- Anthropic’s Opus 5 can now error out when thinking is off but max effort is on — AlexKim · 2026-07-28
- Claude Opus 5 ships at the same price and nears Fable 5 on coding — AlexKim · 2026-07-28
- [source] Claude Opus 5 keeps Opus 4.8 pricing while matching Fable 5 within 0.5% on coding — AlexKim · 2026-07-28
- Anthropic’s Opus 5 now supports all five effort levels and thinking defaults on 4.8 — AlexKim · 2026-07-28
- [source] Anthropic says Claude Opus 5 is best for complex agentic coding and code review — majidmanzarpour · 2026-07-28