Claude Opus 5 launch: Anthropic brings near-Fable 5 performance at half the token cost

Anthropic has launched Claude Opus 5, its latest flagship model, claiming it delivers performance close to the premium Claude Fable 5 at half the token price. The model is now available as the default option on Claude Max and the most capable model on Claude Pro. With a 1-million-token context window and a $5 per million input token rate, Opus 5 gives Indian AI developers and startups a more cost-effective path to state-of-the-art reasoning and coding.

Pricing and effort settings offer better value

Opus 5 retains the same token rates as its predecessor Opus 4.8: $5 per million input tokens and $25 per million output tokens. A new Fast Mode boosts speed by 2.5x but doubles the price. Anthropic also introduces five effort settings — low, medium, high, xhigh, and max — that let users trade off performance against token consumption. The company recommends using the low and medium settings broadly, as they deliver good results with a fraction of the token use and latency. For coding and agentic tasks, xhigh remains the recommended starting point.

Token efficiency matters because earlier models like Opus 4.7 ended up costing 30 to 40 percent more per task than Opus 4.6 despite identical base rates. Anthropic says Opus 5 offers better value than its predecessor at every effort level, which is a key consideration for Indian enterprises managing AI budgets.

Benchmark performance: coding and knowledge work lead

On Frontier-Bench v0.1, which measures agentic terminal coding, Opus 5 scored 43.3 percent, beating Fable 5 (33.7 percent), GPT-5.6 Sol (34.4 percent), and Opus 4.8 (21.1 percent) by wide margins. On the knowledge work benchmark GDPval-AA v2, Opus 5 achieved an Elo score of 1,861, ahead of Fable 5 (1,747) and GPT-5.6 Sol (1,736). The model also posted a surprising 30.2 percent on ARC-AGI-3, a benchmark for novel problem-solving without memorised patterns. Opus 4.8 scored only 1.5 percent, and GPT-5.6 Sol managed 7.8 percent — a nearly four-fold gap. No Fable 5 result was available for that benchmark.

However, Opus 5 does not win every test. On agentic coding via DeepSWE v1.1, GPT-5.6 Sol leads with 72.7 percent, followed by Fable 5 at 69.7 percent and Opus 5 at 68.8 percent. On health and legal tasks, Fable 5 and Mythos 5 outperform Opus 5 respectively. Anthropic deliberately did not train Opus 5 on cybersecurity tasks, so it falls behind Mythos 5 on exploit generation.

Datos clave

Model	Input token price	Output token price	Frontier-Bench v0.1 score
Claude Opus 5	$5 per million	$25 per million	3%
Claude Fable 5	$10 per million (estimated)	$50 per million (estimated)	7%
GPT-5.6 Sol	$15 per million	$60 per million	4%

Real-world examples and safety features

Anthropic highlights Opus 5’s ability to check its own work and improve iteratively. In one test, the model received a drawing of a machine part and had to create a 3D model in FreeCAD without directly viewing the drawing. Opus 5 wrote its own computer vision pipeline to extract geometry from pixels and reconstructed the part. No other model solved the task after five attempts. In another case, Opus 5 fixed a real bug in a popular open-source package manager, addressing an edge case the community patch had missed — while a competing model had only fixed the surface symptom.

Safety classifiers on Opus 5 trigger about 85 percent less often than on Fable 5, which had drawn heavy criticism for frequent interventions. Opus 5 supports source code vulnerability research but blocks binary-based scanning, penetration testing, and exploit generation. For blocked requests, Claude.ai, Claude Code, and Claude Cowork fall back to Opus 4.8.

Why this matters for Indian AI users

Indian startups and enterprises often face tight budgets for large language model usage. Opus 5 offers a near-flagship reasoning and coding capability at a price point that is more sustainable for high-volume tasks. The ability to tune effort settings also helps teams control costs without sacrificing quality. As AI adoption accelerates in India — from fintech to legal tech to software development — models that deliver strong performance at lower token rates become critical for scaling.

Source: The Decoder at https://the-decoder.com/anthropic-claims-its-new-claude-opus-5-delivers-near-fable-5-performance-at-half-the-token-price/