Claude Opus 5: Frontier-Like Capability Without a Frontier Price
Claude Opus 5 is Anthropic’s new large language model that combines frontier-level coding and knowledge capabilities with stable pricing of USD 5 (approx. RM23) per million input tokens and USD 25 (approx. RM115) per million output tokens, making it the default for Claude Max, the strongest option on Claude Pro, and a central choice for developers through the Claude API. Anthropic has launched Claude Opus 5 across all of its platforms, keeping the same base API pricing as its Opus 4.8 predecessor while claiming substantial gains across coding, knowledge work, computer use and scientific research. The headline is not the model’s raw intelligence but its pricing discipline. In a market where every new LLM generation tends to arrive with a steeper bill, Anthropic’s decision to match Opus 4.8’s rates looks like a direct challenge to rivals on the cost-per-capability front. Rather than invent a new "premium" tier, the company is betting that enterprises are more sensitive to total task cost than to model branding. That is a quietly radical stance: it turns Opus 5 from a luxury option into a new baseline for serious work.
Pricing That Rewards Serious Work, Not Occasional Prompts
Claude Opus 5 pricing matters because it targets sustained coding and knowledge workflows instead of casual chat. The model costs USD 5 (approx. RM23) per million input tokens and USD 25 (approx. RM115) per million output tokens, matching Opus 4.8’s rates while delivering higher scores on internal benchmarks like Frontier-Bench. That means existing budgets can, in theory, buy more solved tasks rather than more tokens. Crucially, Anthropic adds a fast mode that runs Opus 5 at about 2.5 times the default speed for twice the base price, giving teams an explicit way to trade latency against spend for time-sensitive applications. Fast mode is not marketing gloss; it formalizes a tier for live trading systems, incident response tools, and interactive coding environments that cannot tolerate lag. The company also introduces an effort setting, allowing customers to pay more tokens only when deeper reasoning is needed. "Opus 5 at its maximum effort setting reportedly came within 0.5% of Fable 5’s peak result while costing half as much per task," one source notes. For enterprises, that is the kind of quotable number procurement teams care about.
Claude Max and Pro: Opus 5 as the New Everyday Heavyweight
By making Opus 5 the default model on Claude Max and the strongest option on Claude Pro, Anthropic is collapsing the gap between "developer-grade" and "subscription-grade" capability. This is a clear signal: the company expects its most serious users to be subscribers who run daily coding and knowledge work, not only API customers. For Claude Max subscribers, the upgrade is automatic. They now get a model built for daily coding and knowledge work that is designed to approach Claude Fable 5’s intelligence at half the price. In practice, that means better long-running coding tasks, more reliable debugging, and stronger performance on business workflow benchmarks without a plan change. Claude Pro users, meanwhile, gain access to Opus 5 as their strongest available model, which is especially meaningful for teams that rely on browser-based agents and document-heavy analysis. The strategic message is blunt: paying for Max or Pro is no longer a compromise compared with "true" enterprise models. Instead, those tiers become the proving ground for agentic workflows and complex knowledge tasks that later scale through the API.
Claude API Cost and Effort Controls: A New Kind of LLM Tuning
For developers, the Claude API cost story around Opus 5 is less about headline rates and more about predictable knobs. Anthropic keeps the base pricing at USD 5 (approx. RM23) per million input tokens and USD 25 (approx. RM115) per million output tokens, with fast mode doubling the price for roughly 2.5x speed. That is straightforward on paper, but the effort setting reshapes how teams think about LLM pricing. Instead of always paying for maximum reasoning, API customers can tune effort per request, reserving higher levels for complex coding, financial research, or genomics work. Anthropic claims that Opus 5 produced better performance at a given cost than competing models at its high, xhigh and maximum effort settings. In other words, optimization happens at the task level, not at the contract level. Two beta updates deepen this operational focus: developers can change tools mid-conversation without breaking the prompt cache, and they can route safety-flagged requests to another model through automatic fallbacks. That kind of plumbing rarely makes headlines, but it matters when applications depend on long tool-using sessions that cannot fail silently.
LLM Pricing Comparison: Opus 5 Bets on Cost-Per-Task Leadership
Viewed against the broader LLM market, Claude Opus 5 is less a raw IQ play than a cost-per-task statement. Anthropic positions Opus 5 as an everyday alternative to its more capable Claude Fable 5 model, claiming it approaches Fable 5 performance at half the price. Benchmark data backs at least part of that story: Opus 5 scored 1,861 on GDPval-AA v2, compared with 1,747 for Fable 5, 1,593 for Opus 4.8 and 1,736 for GPT-5.6 Sol. On AutomationBench, which tests end-to-end business workflows, Opus 5 reached a 26% pass rate against 17.4% for Fable 5, 17% for Opus 4.8 and 18.1% for GPT-5.6 Sol. Those numbers matter because they speak in the language of procurement: more workflows completed per dollar. Anthropic further claims that Opus 5 produced better performance at a given cost than competing models across several effort settings. Yet the company is careful not to push the model into risky territory, keeping it behind Mythos 5 in offensive cybersecurity and long-running biological research. The bet is clear: enterprises want powerful models that stay cost-efficient and aligned, not frontier systems that chase every last benchmark point. The conclusion is straightforward. Claude Opus 5 pricing does not try to win the LLM arms race on sheer capability. Instead, it pressures competitors on the metric that matters most for real deployments: how much completed, safe work you get per unit of spend. For AI developers and Claude Max subscribers, that is a welcome shift from "bigger" to "better value."
