Anthropic Opus 5
TrialPlatforms
A large language model from Anthropic updated for coding and general AI tasks.
Why it's here
Placed in Trial: 12 article(s) of evidence from 4 source(s), led by research-stage coverage, with 4 in the last 30 days. Confidence 76%.
Evidence (12)
- 7Hacker News·10/5/2026researchAI Agents Identify Two Room-Temperature Magnetic Semiconductor Candidates
A team using Opus 5.5 AI agents reported designing one magnetic semiconductor candidate and identifying another existing material, first made in 1999, that their calculations suggest matches the desired properties. The work focuses on materials for spintronics and memory research, especially systems between ferromagnets and antiferromagnets that could offer faster switching and better storage performance.
- 8The New Stack·10/1/2026securityAI coding agents exposed 13,000 screenshots in public repos
Glow Labs says AI coding agents unintentionally published more than 13,000 internal screenshots to public GitHub repositories while trying to complete normal code-review tasks. The incident, called PixelLeak, affected over 300 organizations and often went undetected because the images were stored in employees’ personal repositories and were not visible to code- or secret-scanning tools.
- 4Simon Willison·9/27/2026open_sourceTool to detect likely Bluesky reply bots
Simon Willison built a tool that analyzes Bluesky profiles for signs of likely reply bots. It checks patterns such as rapid reply timing, accounts that rarely post original content, and repetitive replies to larger accounts.
- 7The New Stack·9/22/2026model_releaseAnthropic and OpenAI cut model costs with new releases
Anthropic and OpenAI each released new AI models designed to reduce usage costs while maintaining strong performance. Anthropic's Opus 5.5 targets complex work such as coding, while OpenAI's GPT-6 Sol and Luna focus on efficiency and speed in smaller or mid-tier model classes.
- 8Simon Willison·8/27/2026securityClaude Code Auto Mode Hit by Prompt Injection Attack
A prompt injection researcher reported an attack against Anthropic's Claude Code auto mode that can trick the agent into executing malicious code from a zip archive, with the researcher claiming a high success rate. In some cases, the safety system detected the compromise but then blocked the agent’s attempt to clean up the malware process, highlighting a failure mode where protection interferes with remediation. The report argues that unattended coding agents should run inside a sandbox with restricted network access and credentials.
- 8The New Stack·8/27/2026breakthroughSai hits 73% on OSWorld 2.0 benchmark
Simular says its computer agent Sai achieved a 73% success rate on OSWorld 2.0, a benchmark for long, real-world desktop tasks, placing it ahead of reported results for GPT-5.6 Sol and Opus 5. The company says Sai is designed to handle routine workplace work at lower cost by combining frontier and specialist models with desktop, web, API, and code execution capabilities.
- 6Simon Willison·8/23/2026researchAnthropic’s top model faces weak adoption despite revenue growth
A Financial Times report says Anthropic’s annualized revenue rose sharply to about $65 billion in July, with the company expecting Q3 profitability and 6,000 customers spending at least $100,000 a year. The same report suggests its most expensive models are seeing limited adoption, while cheaper alternatives are gaining traction based on Ramp billing data.
- 4Hacker News·8/19/2026model_releaseClaude Code issue sparks criticism over Opus 5.0 behavior
A Hacker News discussion highlights a GitHub issue in Anthropic's Claude Code project that claims Opus 5.0 produces incoherent output. The post drew substantial attention and debate, but the source material is primarily a community reaction rather than a formal product announcement or technical release note.
- 4Hacker News·8/14/2026researchHacker News post questions Opus 5’s coding-agent behavior
A Hacker News post argues that Opus 5 feels less pleasant to use as a coding assistant than Opus 4.7, Opus 4.8, and Fable, despite being more capable in benchmarks. The author suggests benchmark-driven training may favor models that make bold assumptions instead of stopping to ask clarifying questions, which is often undesirable in real-world software work.
- 8Ars Technica AI·7/29/2026securityHAWK post-quantum signature scheme withdrawn after flaw found
HAWK, a post-quantum digital signature scheme under review for possible US standardization, has been withdrawn after a flaw was identified. The issue was found with help from Anthropic's security model during the third NIST evaluation round, which is designed to expose exactly these kinds of weaknesses.
- 6Hacker News·7/27/2026researchBenchmarking Opus 5 on SlopCodeBench
This post reports benchmark results for Opus 5 on SlopCodeBench, a coding-focused evaluation. It discusses how the model performs on agentic programming tasks and what the benchmark reveals about coding quality and failure modes.
- 6Ars Technica AI·7/24/2026model_releaseAnthropic launches Opus 5 with efficiency-focused improvements
Anthropic has released Opus 5, the latest update to its model family that is popular for coding and software development tasks. The report says the upgrade appears more focused on token efficiency than on a major jump in agentic coding capability.