Claude Opus 5 Launch: Half Fable 5's Price, Higher Benchmark Scores in 2026
Claude Opus 5 is the AI model Anthropic released on July 24, 2026. If you use AI at work, this launch is worth a look. The model costs half of Anthropic's own top-tier system, yet it scored higher on most benchmark tests.
Why Claude Opus 5 Matters
Price. Claude Opus 5 costs $5 per million input tokens and $25 per million output tokens. A token is the unit an AI counts as it reads and writes, roughly three-quarters of an English word. Fable 5, the flagship Anthropic shipped on June 9, costs $10 in and $50 out. Opus 5 is exactly half. The backdrop is a sprint: Anthropic has released four Claude 5 models in under two months, and OpenAI shipped GPT-5.6 Sol on July 10 claiming it beat Fable 5 on coding.
Claude Opus 5 Benchmarks vs Fable 5 and GPT-5.6 Sol
Start with agentic coding. On Frontier-Bench, which scores how well a model writes and repairs code on its own, Claude Opus 5 posted 43.3. Fable 5 hit 33.7 and GPT-5.6 Sol 34.4. Reasoning showed a wider gap: 30.2% on ARC-AGI-3 against Sol's 7.8%. On OSWorld 2.0, which tests operating a computer the way a person does, Opus 5 reached 70.6% to Fable 5's 66.1%. It also solved all six IMO 2026 problems for a perfect 42. Fable 5 held one lead: SWE-bench Pro, 80.0% to 79.2%.
So How Does This Affect You?
Performance and price have started to decouple. For two years, frontier-grade capability lived only in the most expensive tier. Claude Opus 5 pulls it into the standard tier, which changes the math for any company budgeting an AI rollout: the same spend now buys roughly twice the work. Competition shifts from raw power to cost per result.
What Comes Next
Analysts expect the price war among frontier models to continue. Opus 5 also carries a 1 million token context window. Still, no single model owns everything. Opus 5 trails Fable 5 slightly on production coding, and Anthropic says Mythos 5 leads on cybersecurity and biology research.
Key Takeaways
① Half the price - Opus 5 runs $5 in and $25 out per million tokens, exactly half of Fable 5.
② Scores flipped - It leads on agentic coding, reasoning, and computer use.
③ New battleground - The contest moves from raw power to cost per result.
Reading the AI race as a pure leaderboard is getting harder. When the question is how much work a dollar buys, tool choices change with it.
👉 Claude Fable 5 Launch: Mythos-Class AI Goes Public, Free Until June 22 - also worth a read.
📌 Sources: Anthropic, Axios, MarkTechPost, Forbes (2026)
댓글
댓글 쓰기