Anthropic's Opus 5 Sets ARC-AGI-3 Record at 30.2%, Beating GPT-5.6 Sol
Anthropic's Claude Opus 5 set a new high on ARC-AGI-3, scoring 30.2% and overtaking the previous best of 7.8% by GPT-5.6 Sol, according to ARC Prize.
The score came a day after Anthropic launched Opus 5 on all paid Claude plans and its API at $5 per million input tokens and $25 per million output tokens, the same price as Opus 4.8 and, by Anthropic's comparison, half the price of Fable 5. Not every benchmark favored the new model: GPT-5.6 Sol still led DeepSWE 72.7% to 68.8%, while a separate benchmark post said Opus 5 improved 20 points over Opus 4.8 on AlmanBench.
From the sources (25 posts)
@koltregaskesClaude Opus 5 has been spotted, release is very soon. Maybe today.
@mark_kOpus 5 is coming today from @AnthropicAI 🔥
@m1astraClaude Opus 5 now appears to be rolling out across providers.
@theoSo, uh, is Opus 5 still happening today?
@m1astraUpdate: Opus 5 is already live for some accounts, served as "Opus 4.8" in Claude web and Claude Code. It knows things 4.8 doesn't and the output gap is clear, especially frontend.
@apples_jimmyRT @M1Astra: Update: Opus 5 is already live for some accounts, served as "Opus 4.8" in Claude web and Claude Code. It knows things 4.8 doe…
@wesrothClaude Opus 5 may already be running for a small number of users without being labeled as Opus 5.
@claudeaiIntroducing Claude Opus 5. It's a thoughtful and proactive model that comes close to the frontier intelligence of Fable 5 at half the price.
@claudeaiOn several coding and knowledge work evaluations, Opus 5 is the new state-of-the-art:
@claudeaiOpus 5 is also highly efficient. It outperforms other models for a similar or lower cost per task:
@claudeaiOn ARC-AGI-3, an evaluation where AI models must solve novel problems, Opus 5’s score is three times as high as the next best model.
@claudeaiAccording to our automated behavioral audit, Opus 5 is our most aligned model to date. Compared to our other models, it shows the lowest rates of reckless or deceptive behavior, and the strongest adherence to Claude’s Constitution. https://
@claudeaiOpus 5 is stronger than Opus 4.8 on cybersecurity tasks. But it remains substantially behind Mythos 5 at developing exploits. Its safeguards are designed to allow developers to identify and fix software vulnerabilities, while blocking high
@claudeaiOpus 5 is available today on all paid plans and the Claude API, priced the same as Opus 4.8. It’s the default model on Claude Max, and the strongest on Claude Pro. It’s also offered in Fast mode, which runs around 2.5× the default speed. R
@anthropicaiRT @claudeai: Introducing Claude Opus 5. It's a thoughtful and proactive model that comes close to the frontier intelligence of Fable 5 at…
@andrewcurran_Opus 5 is live for me right now.
@firstsquawkANTHROPIC LAUNCHES OPUS 5 AI MODEL WITH FABLE 5 SIMILAR FEATURES AT 50% LOWER COST, CLAIMS COMPANY.
@yahoofinanceClaude's latest model, Opus 5, has been released by Anthropic.
@financialjuiceAnthropic unveils Opus 5 AI model with Fable 5-style features at half the cost: company
@mlstreettalk👀👀👀
@danshipperBREAKING: Claude Opus 5 is OUT NOW! And…it’s a hard model to love. We’ve spent the last week @every testing it across coding, writing, knowledge work, and our internal agent. It argued with instructions, stopped before the work was finis
@mweinbachDAMN OPUS
@cursor_aiClaude Opus 5 is now available in Cursor! It matches Fable 5 on CursorBench (66.7 vs 66.5 at default effort) at half the price. Unlike Fable, it's also compatible with Zero Data Retention.
@cnbcAnthropic's new AI model rivals Fable 5 and is cheaper as businesses fret about costs
@andrewcurran_Benchmarks. As many predicted, including me, it is better than Fable in many categories. It had to be in this environment.