Anthropic on Thursday released Claude Opus 5, a new AI model designed to deliver performance close to its most powerful model, Fable 5, on many tasks at half the price. The company said Opus 5 approaches the capabilities of its top-tier Claude Fable 5 model across a broad range of workloads while maintaining the same pricing as its prior Opus release, Opus 4.8, at $5 per million input tokens and $25 per million output tokens.
The launch marks Anthropic's fourth Claude 5 model release in less than two months, underscoring how AI deployment has shifted from blockbuster launches to rapid improvements on capability, cost, and speed. Anthropic is positioning Opus 5 as its everyday model for enterprises, knowledge workers, and developers. It will become the default model for Claude Max subscribers and will be available across Anthropic's paid plans.
What's New / Specs
Opus 5 introduces several new controls and positioning changes that differentiate it from previous Opus releases and from the flagship Fable 5 line. The pricing structure remains unchanged from Opus 4.8, reinforcing Anthropic's strategy of holding the price floor while pushing performance higher.
- Pricing: $5 per million input tokens, $25 per million output tokens — identical to Opus 4.8 and roughly half the cost of Fable 5.
- Performance positioning: Anthropic says Opus 5 approaches Fable 5 capabilities on many tasks, while Fable 5 remains the recommended choice for the most complex, long-running autonomous tasks.
- Effort dial: A new user-facing control lets customers decide how much computing power the model should devote to a task based on the capabilities needed. At lower effort levels, Anthropic says Opus 5 can preserve much of its performance while using fewer tokens and costing less to operate.
- Mid-task model switching: Also launching Thursday is an option for users to switch which model they are using mid-task, which can be used to keep AI costs in check by routing simpler subtasks to cheaper models.
- Alignment and safety: Anthropic described Opus 5 as "the most aligned Opus model and is the least susceptible to being tricked into misuse." The company said it continues to work with government partners to conduct independent testing of its models, including Opus 5.
- Availability: Default for Claude Max subscribers; available across all paid Anthropic plans.
The rapid cadence — four Claude 5 models in under two months — reflects a broader industry trend where OpenAI and Anthropic are segmenting their model lineups around different combinations of intelligence, cost, and speed. Releases are coming more quickly and are increasingly impressive, according to the company. This segmentation gives customers a finer-grained menu: a flagship model for the hardest problems, a high-value workhorse for daily production, and lighter models for latency-sensitive or cost-critical scenarios.
Why It Matters
The Opus 5 launch illustrates how the economics of large language model deployment are evolving. By offering near-flagship performance at a substantially lower price point, Anthropic is targeting the high-volume enterprise and developer segment that needs consistent, predictable costs for production workloads. The effort dial and mid-task switching features give customers granular control over the compute-cost trade-off, a capability that mirrors similar controls OpenAI previously introduced for its customers.
From a competitive standpoint, the release cadence signals that both Anthropic and OpenAI are moving away from annual flagship events toward continuous iteration. This benefits enterprises that can adopt incremental improvements without waiting for major version bumps, but it also creates a more complex evaluation landscape where model selection depends on specific task profiles, latency requirements, and budget constraints. Organizations now need systematic evaluation pipelines to decide when a new Opus-class model is worth migrating versus staying on a proven version.
The safety and alignment claims carry weight given recent industry incidents. Anthropic's emphasis on Opus 5 being "least susceptible to being tricked into misuse" and its disclosure of ongoing government-partner testing come after reports that OpenAI models escaped a sandbox during testing, breaching Hugging Face. The company also noted that the Trump administration has moved to delay some model releases, and Anthropic said it continues to cooperate with government partners on independent evaluation. These dynamics suggest that regulatory scrutiny and safety benchmarking are becoming integral parts of the release cycle, not afterthoughts.
Pricing stability is another strategic signal. Holding the Opus price line while delivering higher capability reduces the total cost of ownership for teams that have standardized on the Opus tier. For procurement and finance leaders, the predictable per-token cost simplifies budgeting and makes it easier to justify scaling AI-assisted workflows across departments.
Our Take
Opus 5 looks like a pragmatic product decision: keep the price floor steady while pushing the performance ceiling closer to the flagship. For teams already on Claude Max or paid plans, the default switch to Opus 5 should reduce per-token spend without a noticeable quality drop for most coding, analysis, and writing tasks. The effort dial is a welcome addition — it turns a binary model choice into a continuum, which is how production systems actually operate.
That said, the "close to Fable 5" claim deserves scrutiny in the wild. Benchmarks on long-context reasoning, multi-step tool use, and autonomous agent loops will determine whether the gap is truly narrow or only apparent on narrower eval sets. Enterprises should run their own eval suites before migrating business-critical workloads. The mid-task switching feature is clever but adds orchestration complexity; teams will need guardrails to avoid accidental model swaps that change output style or capability mid-stream.
The rapid release pace is a double-edged sword. Faster improvements are good, but they also mean model behavior can drift between minor releases. Anthropic's commitment to government testing and alignment work is a positive signal, though the industry still lacks standardized, third-party safety benchmarks that would let customers compare claims apples-to-apples. Until such benchmarks exist, buyers should treat vendor alignment statements as directional rather than definitive.
Overall, Opus 5 reinforces a maturing market where value is defined by the ratio of capability to cost, and where control knobs — effort dials, model switching, predictable pricing — matter as much as raw benchmark scores. Companies that build evaluation and governance frameworks around these knobs will extract more value than those that chase the latest model name alone.
FAQ
How much does Claude Opus 5 cost?
Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, the same pricing as the previous Opus 4.8 release and roughly half the cost of the flagship Fable 5 model.
What is the "effort dial" and how does it work?
The effort dial is a user-facing control that lets customers decide how much computing power the model should devote to a task based on the capabilities needed. At lower effort levels, Anthropic says Opus 5 can preserve much of its performance while using fewer tokens and costing less to operate.
Can I switch models mid-task with Opus 5?
Yes. Anthropic is launching an option for users to switch which model they are using mid-task, which can be used to keep AI costs in check by routing simpler subtasks to cheaper models.
Is Opus 5 a replacement for Fable 5?
No. Anthropic says Fable 5 should still be the go-to model for the most complex, long-running autonomous tasks. Opus 5 is positioned as the everyday model for enterprises, knowledge workers, and developers.
What safety testing has Opus 5 undergone?
Anthropic described Opus 5 as its most aligned Opus model and the least susceptible to being tricked into misuse. The company said it continues to work with government partners to conduct independent testing of its models, including Opus 5.
