Claude
Anthropic's AI assistant — Opus 5.5, Fable 5.1 and Sonnet 5 — with best-in-class writing, code review, a 1M-token context and Claude Code.
About Claude
Tags
Reviews (6)
Claude Haiku 5.5 Review: Luna’s Price, a Higher Ceiling
Claude Haiku 5.5 scores 43 on the Artificial Analysis Intelligence Index at maximum effort, 26 points above Haiku 4.5, at GPT-6 Luna's $0.10/$0.50. Up to a score of 38, Luna gets there for the same money or less. Above it, Luna has nothing to match Haiku 5.5, which also streams its answers faster.
- +Scores 43 on the Artificial Analysis Intelligence Index at max — 26 points above Haiku 4.5
- +Costs less per index task than Haiku 4.5 at every setting while scoring at least 12 points more
- +Reaches 41 and 43 at xhigh and max — scores GPT-6 Luna does not reach at any setting
- -Up to a score of 38 GPT-6 Luna gets there for the same money or less
- -Took about 13 seconds to a first token at low and medium on its first day of measurements
- -Prompts over 100K tokens cost five times as much per token
Claude Sonnet 5.5 Review: Great Value Until You Max It
Claude Sonnet 5.5 scores 56 on the Artificial Analysis Intelligence Index, second only to Claude Opus 5.5, at Sonnet 5's unchanged $2/$10. Up to high effort it is cheap and fast, answering in seconds where Opus 5.5 takes half a minute. At maximum effort it costs more per task than Opus 5.5 for a lower score.
- +Second on the Artificial Analysis Intelligence Index at 56 — behind only Claude Opus 5.5
- +Same $2/$10 list price as Sonnet 5; cache reads halved to $0.10 on 7 October
- +At medium effort beats Sonnet 5’s best score for about a ninth of the cost per task
- -Max effort costs $7.60 a task — more than Opus 5.5 at max for a lower score
- -Thought for 32 to 48 minutes on each of our build briefs at max effort
- -Generated 410M output tokens on the index at max — nearly five times the median
Claude Opus 5.5 Review: The New No. 1 Costs Less
Claude Opus 5.5 takes sole first place on the Artificial Analysis Intelligence Index, scoring 58 to the 53 shared by Claude Fable 5.1 and GPT-6 Astra. It lists at $4/$20, a fifth under Opus 5. The better story sits one effort level down: at high it scores 54, above every rival, for $1.82 a task. It also won our three-round build-off, and it was the slowest entrant in every round.
- +First of 212 models on Artificial Analysis Intelligence Index v4.3.2 at 58
- +At high effort scores 54 for $1.82 a task — above every rival at any setting
- +Beats Claude Fable 5.1 by five index points for 22% less per task
- -At max effort it emits 260M output tokens on the index — about three times the median
- -Max effort costs $5.98 a task — no cheaper than Opus 5 and nearly twice GPT-6 Astra
- -Slowest entrant in every round of our build-off at 59 to 77 minutes a brief
Claude Opus 5 Review: Smartest, and Most Annoying
Opus 5 tops the Artificial Analysis Intelligence Index at 61 and posted a near-4x jump on ARC-AGI-3, all at the same price as Opus 4.8. Then the reviewers who tested it called it 'brilliant but annoying': it argues with instructions, stops early, does unrequested refactors, and breaks prompt scaffolding tuned for Opus 4.8. One more finding deserves attention — a reported jump in its hallucination rate.
- +Ranked first on the Artificial Analysis Intelligence Index at 61
- +ARC-AGI-3 leap to 30.2% from a 7.8% prior best
- +Within 0.5% of Fable 5 on CursorBench at half the cost
- -Argues with instructions and stops before finishing work
- -Reported hallucination rate up 14 points to 50%
- -Verbose by default — confirmed in Anthropic's own docs
Claude Fable 5 Review: Most Capable, at a Premium
Anthropic's Claude Fable 5 is the most capable model the public can use today — topping SWE-Bench Pro and excelling at vision and long tasks. It is also the priciest major model and ships with hard safety guardrails. Our first-look verdict.
- +Best agentic-coding result to date (80.3 on SWE-Bench Pro)
- +State-of-the-art vision plus stronger long-horizon autonomy
- +Mythos-class capability finally available to everyone
- -Most expensive major model at $10 in / $50 out per million tokens
- -Safety classifiers can block legitimate security research
- -Mandatory 30-day retention even on zero-retention plans
Claude Review 2026: Plans, Models and Verdict
Anthropic turned the lineup over again: Fable 5.1 replaced Fable 5 at the top on September 1, Opus 5 replaced Opus 4.8 at the same price, and Sonnet 5’s intro pricing has ended. Where Claude leads, where it still struggles, and whether Pro and Max are worth it in September 2026.
- +Best-in-class nuanced writing
- +Fable 5.1 leads long-horizon agentic coding (55.8% Terminal-Bench 4.0)
- +Opus 5 tops the Artificial Analysis index at Opus 4.8’s price
- -Fable 5.1 is the most expensive frontier model per task ($10/$50 list and heavy token use)
- -Opus 5 argues with instructions and needs prompt re-tuning
- -Hard safety guardrails still block some legitimate cyber/bio work
Similar Tools
ChatGPT
Text & ChatOpenAI's ChatGPT assistant — conversation, writing, coding, image generation and data analysis, now on the GPT-6 family: Astra, Sol and Luna.
Perplexity
Text & ChatAI-powered search engine that answers questions with cited sources. A smarter alternative to Google.
Gemini
Text & ChatGoogle's AI model with live internet access, image analysis, and deep Google service integration.