Claude Fable 5.2 API Integration: Model ID, Specs and Access Checks

The official Claude API model ID for Claude Fable 5.2 is claude-fable-5-2. Anthropic lists it for demanding reasoning and long-horizon agentic work, with a 1M-token context window and a 128K-token maximum output.

For LLMAPI integration, verify access separately: copy the request ID from the live catalog for your billing mode, confirm Anthropic-protocol support, and use the matching key. An official model ID is not a substitute for checking your LLMAPI catalog and key permissions.

Official model ID and specifications

The Claude models overview lists these Fable 5.2 specifications:

Setting	Official listing
Claude API ID / alias	claude-fable-5-2
Intended use	Demanding reasoning and long-horizon agentic work
Input → output	Text and images → text
Context window	1M tokens
Maximum output	128K tokens
Thinking	Adaptive, always on
Default effort	high
Comparative latency	Moderate
Reliable knowledge cutoff	June 2026

Anthropic recommends starting with Opus 5.5 for most workloads and considering Fable 5.2 for demanding reasoning or when evaluations on Opus 5.5 at higher effort still fall short. Use that as a model-selection starting point, not as a measured performance claim for your application.

Keep model-specific settings separate. The Opus 5.5 overview documents a 300K-output Message Batches beta for Opus 5.5. Do not use that Opus-specific setting as the basis for a Fable 5.2 configuration; Fable’s listed maximum output is 128K.

For official pricing, consult the current models overview. For LLMAPI costs, use the plans and API billing page and the selected channel’s current pricing rather than treating vendor pricing as your LLMAPI rate.

Verify LLMAPI catalog access and billing mode

The LLMAPI model-selection guide makes the live catalog and your key permissions authoritative for request IDs, protocols and features. Before configuring Fable 5.2, check all of the following:

1. Catalog entry: Find the intended model in the subscription model list or pay-as-you-go channel catalog. Copy its request ID, not its display name.
2. Protocol: Confirm that the selected entry supports the protocol used by your client. For Claude Code, check Anthropic-protocol support.
3. Billing mode: Identify whether the entry belongs to subscription coverage or metered API billing.
4. Key and allowance: Use a subscription key with the plan allowance, or an API billing key with the separate API balance.
5. Permissions: Check that your selected key is permitted to use that entry and its required features.

The subscription model list is synchronized with GET /v1/models. Use that list when checking subscription coverage; do not treat it as a substitute for checking pay-as-you-go channels.

Subscription and metered API billing are separate. Topping up API credit does not increase subscription quota. Likewise, a client’s OpenAI-protocol compatibility does not mean every model is included in your subscription.

If you cannot find the intended Fable 5.2 entry for your billing mode, stop before substituting a similarly named model or assuming the official ID is accepted through that route.

Configure Claude Code’s Anthropic connection

The Claude Code setup guide specifies https://llmapi.pro as the Anthropic base URL, without /v1. The client calls /v1/messages.

The following documented environment setup configures the connection; it does not select or verify Fable 5.2 access. Replace YOUR_API_KEY with the LLMAPI key matching your chosen billing mode.

For macOS or Linux:

export ANTHROPIC_BASE_URL=https://llmapi.pro
export ANTHROPIC_AUTH_TOKEN=YOUR_API_KEY

For Windows PowerShell:

$env:ANTHROPIC_BASE_URL="https://llmapi.pro"
$env:ANTHROPIC_AUTH_TOKEN="YOUR_API_KEY"

The setup guide uses claude --model to select a model, but its examples name claude-opus-5 and claude-fable-5. Those are not the official Fable 5.2 ID shown above. For your model selection, copy the exact request ID confirmed in your account’s live catalog rather than reusing the guide’s Fable 5 example unchanged.

Keep the protocol and base URL paired correctly. The LLMAPI documentation distinguishes the Anthropic base URL, https://llmapi.pro, from the OpenAI-compatible base URL, https://llmapi.pro/v1. For an OpenAI-compatible client, check the selected model’s protocol support before using that route.

Check capabilities before choosing request settings

Anthropic’s models overview documents a Models API response containing max_input_tokens, max_tokens and a capabilities object for available models. Use these fields when checking official API limits and supported settings rather than inferring capabilities from the model name.

Fable 5.2 is listed with adaptive thinking always on. Do not build its configuration around an assumption that thinking can be disabled. The official Models API also exposes thinking.types.disabled: its supported value indicates whether the model accepts the disabled-thinking setting, with additional request restrictions still possible even when that value is true.

For LLMAPI, verify features against the selected catalog entry and key permissions separately. Treat the official specifications as vendor specifications, not as a promise that every LLMAPI route exposes identical settings.

Make a small account-specific access check

After confirming the catalog entry, protocol and billing key, make a small request through your configured client before attempting a long-context or agentic workload. For example, use a generic prompt such as “Reply with one short sentence.” This is an illustrative check, not a reported Fable 5.2 test result.

Record the selected request ID, billing mode and whether the request succeeds. If it fails, first recheck the exact catalog ID, the key type and permissions, and the protocol/base-URL pairing. Do not infer access from a display name or a client menu entry alone.

Keep the initial check narrow: a successful short request is the first integration check, not your evaluation of demanding reasoning, long-horizon behavior or maximum-context use. Evaluate those requirements separately before choosing Fable 5.2 for the workload.