Applied artificial intelligence, neural systems, and smart agents.
Fusing generative AI, multi-agent frameworks, and high-throughput data processing to automate business workflows.
Claude Fable 5
Fable 5 went into production on 9 June 2026 and, at a $10 in / $50 out rate, is the highest-cost model in general circulation on Anthropic's production line. It holds rank one on the Arena text leaderboard, but ranks second in this reference's coding table, behind a model billed at half that rate.
- rank one on the Arena text board
- 1M token context window
- knowledge cutoff January 2026
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. Change log
Why the highest-cost Anthropic model ranks second in the coding table
On WebDev Arena, Fable 5 logs a score of 1628 and Opus 5 logs 1692. A sixty-four-point gap, in favour of the model billed at half the output rate. On the general text leaderboard, the order inverts: Fable 5 leads at 1506 and Opus 5 sits at rank seven with 1494. There, the gap is only twelve points.
So the coding gap runs more than five times the conversation gap, and points the opposite direction. The workload split is derived directly from those two figures: route code to Opus 5, route long-form text and conversation to Fable 5. Running both workload types on a single model means overpaying on one of the two.
Its knowledge window trails the cheaper option
Fable 5's knowledge cutoff is fixed at January 2026, against May 2026 for Opus 5. The model billed at double the rate trails by four months, a gap the pricing table shows no trace of. On a framework that shipped changes this year, that gap surfaces in the first two or three outputs, not in a benchmark.
The likely cause is release sequencing: Fable 5 went into production in June, Opus 5 in July. For anyone choosing today, the cause changes nothing, and the operational outcome changes plenty.
The field where Fable 5 scores zero
This reference's coding category carries a score-per-dollar field that divides the leaderboard rating by the output rate. Fable 5's value comes out at 32.6, the lowest of every candidate with a verified rate, so it draws nothing from that field's twenty points of weight. DeepSeek V4 Flash logs 5660 in the same field. The arithmetic is open on the category page and independently verifiable.
Do not read that as an argument for the lowest rate available. Fable 5 takes full marks in the tooling-maturity field and is deployed across five delivery routes, from Anthropic's own API to Claude Platform on AWS and Microsoft Foundry. If your organization's infrastructure already runs a contract on one of those clouds, that field carries more weight for you than the rate does.
Specifications
Every number here comes from the vendor's own page, with that page linked beside it.
- API model ID
- claude-fable-5 Source
- Context window
- 1 million tokens Source
- Max output
- 128k tokens Source
- Input price
- $10 per million tokens Source
- Output price
- $50 per million tokens Source
- Cached input price
- $1 per million tokens Source
- Knowledge cutoff
- January 2026 Source
- Where it ships
- Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, Microsoft Foundry Source
Using it from Iran
This column carries a date and says how it was checked, because most listicles guess it. Where we have only read a vendor policy page, the note below says exactly that.
- Reachable
- blocked
- Payment
- no working route
- Free tier
- no
- Measured on
How we checked: Iran is not registered on Anthropic supported-countries list. Extracted from Anthropic own policy document, not from a network measurement.
Documented strengths
- Rank one on the Arena text leaderboard at a score of 1506, ahead of Opus 5
- A one-million-token context window with a 128k-token output ceiling
- Coverage across five delivery routes: the Anthropic API, Bedrock, Claude Platform on AWS, Google Cloud and Microsoft Foundry
- Full marks in the tooling-maturity field of the coding table
Room for improvement and technical considerations
- It carries twice the price tier of Opus 5 and still sits below Opus 5 on the web-development leaderboard.
- Its knowledge cutoff is January 2026, four months behind Opus 5.
- It finishes at the bottom of this reference's score-per-dollar field and draws none of that field's twenty points of weight.
- No official deployment route from Iran: it is absent from Anthropic's supported-countries list and Iranian card payment processing is not operational.
- Anthropic published no absolute benchmark values for this model either, leaving part of the table fields empty for it.
Technical verdict
If you want a single model for the entire workload and cost is not the first metric, Fable 5 is a defensible pick. But a team spending most of its time on coding pays double with Fable 5 to score lower on that exact workload. In this reference, that condition routes to Opus 5, with Fable 5 held back for writing and conversation.
Frequently raised questions, with documented answers
From an architecture standpoint, does Fable 5 lead Opus 5
It depends on the workload, and both answers carry a supporting figure. On WebDev Arena, Opus 5 scores 1692 against Fable 5's 1628, at half the rate. On the general text leaderboard, Fable 5 leads at 1506 and Opus 5 sits at rank seven with 1494.
What is Fable 5's pricing architecture
Ten dollars per million input tokens and fifty per million output tokens. That is API-layer pricing, and it is separate from the monthly Claude app subscription, which is billed independently.
Why does the higher-rate model rank lower on coding workloads
Because a higher rate does not necessarily buy a higher score. Anthropic positions Fable 5 as its most capable model in general circulation, and the general text board agrees, but Opus 5 went into production four months later and passed it on the web-development board.
What is the access path from Iran to Fable 5
No official route exists. Iran is not registered on Anthropic's supported-countries list, so there is no legitimate signup path and no Iranian-card payment processing. This reference does not sell accounts and does not propose circumvention methods.