Anthropic Ships Claude Opus 5 on 24 July 2026
Anthropic released Claude Opus 5 on 24 July 2026, making it the fourth model the company has shipped in under two months, following Mythos 5, Fable 5, and Sonnet 5 in June. The model is available immediately via the Anthropic API under the identifier claude-opus-5 and is the default model on Claude Max subscriptions. Anthropic's positioning is precise: Opus 5 delivers near-Fable-5 intelligence at roughly half the price of the company's current top-tier model, with no breaking changes from Opus 4.8.
Benchmark Performance: How Opus 5 Stacks Up
The headline figure from Anthropic's launch is a score of 43.3 percent on Frontier-Bench v0.1, an industry evaluation covering reasoning, knowledge work, and long-context instruction-following. This compares with 33.7 percent for Fable 5 and 18.7 percent for Opus 4.8 on the same benchmark — placing Opus 5 clearly above Anthropic's previous frontier model in this tier. On OSWorld 2.0, a benchmark for computer-use tasks that require the model to operate software interfaces autonomously, Opus 5 surpassed Fable 5's top result at approximately one-third of Fable 5's inference cost per task. Anthropic also reports new state-of-the-art performance on agentic-coding benchmarks. The one area where Anthropic acknowledges Opus 5 remains behind is cybersecurity tasks, where Mythos 5 — the company's most capable research model — continues to lead.
Pricing and Context Window
Anthropic priced Opus 5 at five dollars per million input tokens and twenty-five dollars per million output tokens, identical to the Opus 4.8 pricing. Teams running Opus 4.8 in production receive the capability jump with no increase in spend simply by updating the model identifier. A Fast mode option doubles the rate to ten dollars per million input tokens and fifty dollars per million output tokens, delivering approximately 2.5 times the throughput for latency-sensitive applications. Opus 5 ships with a one-million-token context window, matching Fable 5 and suitable for processing entire codebases, lengthy documents, or extended multi-turn agentic sessions.
The Effort Toggle
A feature introduced with Opus 5 is a low, medium, and high effort toggle that allows teams to calibrate how much reasoning compute the model applies per request. At low effort, Opus 5 responds more quickly and cheaply, trading some accuracy for speed — useful when approximate answers are acceptable. At high effort, the model applies extended reasoning, improving accuracy on complex technical, mathematical, or multi-step tasks at the cost of higher latency and marginally higher token usage. The toggle enables product teams to tune cost and quality dynamically within a single deployment, rather than routing requests between different model tiers.
Agentic-Coding and Computer Use Improvements
Opus 5 ships with specific improvements to two agentic workloads: software development and computer use. On agentic-coding benchmarks, Anthropic reports new state-of-the-art performance, meaning Opus 5 plans and executes multi-file code changes, runs tests, interprets results, and iterates with greater reliability than its predecessors. On computer-use tasks, the OSWorld 2.0 results show a concrete advancement over Fable 5 in both accuracy and cost efficiency — relevant for teams building browser automation, workflow orchestration, and enterprise-grade AI agents that need to operate real interfaces.
What Opus 5 Means for AI Product Teams in India
For Indian enterprises evaluating Anthropic's model tier, Opus 5 clarifies the cost structure substantially. Fable 5 and Mythos 5 remain the right choice for the highest-stakes agentic deployments where cost is secondary. Opus 5 fills the position that until now had no ideal occupant: complex, long-running agentic tasks where Sonnet 5 is underpowered but paying for Fable 5 is disproportionate. Indian software product companies building AI coding assistants, document automation platforms, or enterprise agentic workflows will find the combination of a one-million-token context window, effort toggle, and near-Fable-5 benchmark performance the most cost-effective path to frontier-tier agentic capability in July 2026. For teams already running Opus 4.8, the migration is a single model identifier change — the pricing, API contract, and output format are unchanged.
The Bottom Line
Anthropic released Claude Opus 5 on 24 July 2026. On Frontier-Bench v0.1, Opus 5 scores 43.3 percent — above Fable 5's 33.7 percent and ahead of Opus 4.8's 18.7 percent. Pricing is unchanged at five dollars per million input tokens and twenty-five dollars per million output tokens, with Fast mode at ten and fifty dollars per million tokens at 2.5 times throughput. The model ships with a one-million-token context window, a low-medium-high effort toggle, and new state-of-the-art results on agentic-coding and computer-use benchmarks. For Indian AI product teams, Opus 5 is the clearest argument yet to run production agentic workflows at frontier quality without a corresponding increase in per-token spend.
Frequently Asked Questions
What is Claude Opus 5 and when was it launched?+
Claude Opus 5 is Anthropic's latest large language model, released on 24 July 2026. It is available via the Anthropic API as claude-opus-5 and is the default model on Claude Max. Anthropic positions it as delivering near-Fable-5 intelligence at roughly half the price of Fable 5, with no breaking changes from Opus 4.8. It is the fourth model Anthropic shipped in under two months in 2026, following Mythos 5, Fable 5, and Sonnet 5 in June 2026.
What benchmark scores did Claude Opus 5 achieve?+
On Frontier-Bench v0.1, Claude Opus 5 scores 43.3 percent, compared with 33.7 percent for Fable 5 and 18.7 percent for Opus 4.8, placing it above Anthropic's prior frontier model in this tier. On OSWorld 2.0, which evaluates computer-use tasks requiring autonomous operation of software interfaces, Opus 5 surpassed Fable 5's top result at approximately one-third of Fable 5's inference cost. Anthropic also reports new state-of-the-art performance on agentic-coding benchmarks. The one area where Opus 5 trails a more capable model is cybersecurity tasks, where Mythos 5 continues to hold the lead.
What does Claude Opus 5 cost and how does it compare to Opus 4.8?+
Claude Opus 5 is priced at five dollars per million input tokens and twenty-five dollars per million output tokens — identical to Opus 4.8. Teams currently using Opus 4.8 receive the capability upgrade with no cost increase by updating the model identifier to claude-opus-5. A Fast mode is available at ten dollars per million input tokens and fifty dollars per million output tokens, delivering approximately 2.5 times the throughput for latency-sensitive applications. The model ships with a one-million-token context window and a low, medium, and high effort toggle for per-request calibration of reasoning depth and cost.
What is the effort toggle in Claude Opus 5 and how does it affect cost?+
Claude Opus 5 introduces a low, medium, and high effort toggle that controls how much reasoning compute the model applies per request. At low effort the model responds quickly and cheaply, suitable when approximate answers are acceptable. At medium effort — the default — the model applies standard reasoning. At high effort the model applies extended reasoning, improving accuracy on complex tasks such as multi-step code generation and deep document analysis, at the cost of higher latency and marginally more token usage. The toggle lets product teams dynamically tune cost and quality within a single model deployment rather than routing requests between different model tiers.
Written by
TechPillow Team
Sharing insights on technology, product development, and the Indian tech ecosystem.