Anthropic’s Claude Opus 5 scores 43.3% on Frontier-Bench and stays at $5/$25
mark_k · x · 2026-07-25
Claude Opus 5 is being positioned by Anthropic as a workhorse model for serious coding, agents, and professional knowledge work.
- It posts strong scores across frontier and agentic benchmarks, including 43.3% on Frontier-Bench, 30.2% on ARC-AGI-3, 70.6% on OSWorld 2.0, 90.8% on BrowseComp, and 26.0% on AutomationBench.
- The author argues this is less a “higher IQ” model and more an agent reliability upgrade: better at using tools, operating computers, searching, verifying work, and finishing long tasks.
- Pricing stays at $5 / $25 per million tokens, the same as Opus 4.8, with a 1M-token context window.
- It is not a clean sweep: GPT-5.6 Sol still leads DeepSWE, while Fable 5 remains stronger on some legal and health tasks and slightly ahead in some untethered reasoning tests.
Overall, the post frames Opus 5 as a practical upgrade for day-to-day agentic work rather than just another benchmark bump.
Related event: Anthropic Releases Claude Opus 5: SOTA Performance at Half the Price(128 posts)→
More from Models
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- ChatGPT tells user their question is too hard and to 'accept dumber answers' — phido3000 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11