Claude Opus 5

Anthropic launched Claude Opus 5 on July 24, 2026 as a vision-language model positioned as a capable everyday workhorse behind Fable 5. It accepts text and images up to 1 million tokens, outputs text up to 128,000 tokens at about 52.8 tokens per second, and carries a May 2026 knowledge cutoff. Pricing is set at $5 input, $0.50 cached, and $25 output per million tokens, with a fast mode at $10/$1/$50. Opus 5 ships with five reasoning levels (low, medium, high, xhigh, max, defaulting to high), prompt caching from 512 tokens, tool use, and no data retention, and tops Artificial Analysis' Intelligence Index at 61 points while easily leading ARC-AGI-3. It becomes the default for Claude Max subscribers ($100-$200 per month) and the strongest model available to Claude Pro subscribers ($20 per month), though Anthropic disclosed no parameter count, architecture, or training details.

Claude Opus 5 ranks first of 170 models on the Artificial Analysis Intelligence Index, scoring 61 in max reasoning mode versus Fable 5's 60. Released July 24, 2026, it also leads Frontier-Bench by more than double Opus 4.8, triples the next-best result on ARC-AGI 3, and beats Fable 5 on OSWorld 2.0 at roughly one-third the cost. Zapier AutomationBench results run about 1.5 times the nearest competitor. Anthropic's launch disclosure also acknowledges a clear loss: Opus 5 trails Mythos 5 on cybersecurity benchmarks, the one category where the company did not claim the top spot. Despite that gap, the combination of leading intelligence-index score, strong agentic coding wins, and lower per-token pricing makes Opus 5 the new capability-per-dollar benchmark among Anthropic's mid-tier releases.

Claude Opus 5 launched July 24, 2026, framed as a workhorse near Fable 5's frontier intelligence at roughly half the price, shipping at $5 input and $25 output per million tokens. Benchmark highlights include 43.3% on Frontier-Bench v0.1 (versus GPT-5.6 Sol at 34.4% and Fable 5 at 33.7%), 30.2% on ARC-AGI-3, and a 1,861 GDPval-AA v2 Elo for knowledge work. It is Anthropic's fourth Claude model in under two months. On SWE-bench Pro, Opus 5 scored 79.2%, narrowly behind Mythos 5 (80.3%) and Fable 5 (80.0%). CursorBench 3.2 placed it within 0.5 points of Fable 5's peak, while OSWorld 2.0 results beat Fable 5 at one-third the budget and Zapier AutomationBench delivered 1.5 times the throughput of the next-closest model. The release narrative emphasizes capability per dollar across coding and agentic workloads.

Anthropic released Claude Opus 5 on July 24, 2026, rolling it out immediately across claude.ai, Claude Code, Claude Cowork, and the Claude API at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8. The model ships with a 1M-token context window, extended thinking on by default, and a new per-request effort toggle balancing cost against capability. It becomes the default model for Claude Max and the strongest available on Claude Pro. Launch benchmarks include 96.0% on SWE-bench Verified, 79.2% on SWE-bench Pro (within a point of Mythos 5 and Fable 5), 43.3% on Frontier-Bench v0.1, and 30.2% on ARC-AGI-3, roughly three times the next-best result. Anthropic positions Opus 5 as the workhorse tier, recommending Fable 5 only for long-horizon autonomous agents and Sonnet 5 for high-volume scoped tasks.

Anthropic published Claude Opus 5's launch benchmarks on July 24, 2026, positioning the mid-priced model at or near the top of most evaluations. It scored 96.0% on SWE-bench Verified, saturating the standard software-engineering test, and 79.2% on the harder SWE-bench Pro, trailing Mythos 5 (80.3%) and Fable 5 (80.0%) by under a point. The model also reached 43.3% on the agentic Frontier-Bench v0.1, ahead of Fable 5 and GPT-5.6 Sol. On novel-reasoning tests, Opus 5 hit 30.2% on ARC-AGI-3, roughly three times the next-best score, and led knowledge work with a 1,861 Elo on GDPval-AA v2. It also surpassed Fable 5 on OSWorld 2.0 at about one-third the cost and delivered roughly 1.5 times the pass rate of competitors on Zapier AutomationBench. The headline shift is price-performance leadership at $5 per million input tokens, undercutting more expensive flagships.

CruxDigits' July 24, 2026 review confirms Anthropic released Claude Opus 5 on that date, positioned as a model that "gets close to the most powerful system, Claude Fable 5, at half the price" and immediately made the default model on Claude Max and the strongest model on Claude Pro. Pricing is $5 per million input toke The piece is framed for a Dutch/EU AI consulting audience and serves as a derivative roundup rather than original reporting, repeating Anthropic's launch messaging without linking to a primary Anthropic source in the supplied excerpt. Pricing, default-on-Max positioning, and the half-of-Fable-5 framing are consistent w

CodeRabbit's July 24, 2026 review of Claude Opus 5 for AI code review identifies five concrete changes for Opus 5 versus prior versions: effort becomes the primary control with thinking on by default (only disableable at high or lower effort), per-turn task and effort budget changes mid-conversation without invalidatin The review places Opus 5 inside an accelerated Anthropic release cadence — naming four releases (Mythos 5, Fable 5, Sonnet 5, Opus 5) over the prior two months — and benchmarks Opus 5 against earlier Opus versions back to 4.1, 4.5, 4.7, 4.8, and Fable 5 within CodeRabbit's own code-review harness. It frames Opus 5 as "