Claude Fable 5.2 is suited for tasks requiring multi-step analysis, ongoing checks, and complete delivery, such as complex code reviews, in-depth research, and organizing materials into documents, spreadsheets, or presentation content. It accepts text and image inputs, produces text output, and can connect to your execution environment through tool calls.
Point the Base URL to api.acedata.cloud, configure your platform API key and the model ID below, and use your compatible SDK or client.
api.acedata.cloudclaude-fable-5-2import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ACEDATACLOUD_API_KEY"],
base_url="https://api.acedata.cloud/v1",
)
response = client.chat.completions.create(
model="claude-fable-5-2",
messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)Choose an available protocol for this model. OpenAI SDK uses a Base URL ending in /v1; Anthropic SDK uses the root URL. See each guide for protocol-specific parameters, tools and response formats.
Select the full model ID and use the request format for the corresponding endpoint.
Official capacity is a model specification; actual requests are also subject to the selected interface's request body size, output budget, and timeout limits. For long tasks, save results in stages and set verifiable acceptance criteria.
Give the model the goal, materials, and completion criteria together.
Provide relevant code, error logs, and expected behavior, and have the model trace dependencies, suggest possible causes, and provide verification steps. Add test results to the conversation, then check whether the fix meets the original requirements.
Submit materials around the same question, ask it to organize the evidence, disagreements, and items to be confirmed step by step, then produce a reviewable report. The model can help with analysis and writing, but you still need to provide materials or connect retrieval tools to obtain the latest information.
Input screenshots, charts, and text questions together, and ask the model to explain visible content, compare differences, and identify the basis for its conclusions. When precise numbers are involved, provide the original data and verify image recognition results.
Adaptive thinking remains enabled, and the output budget covers both thinking and the main response.
Adjust reasoning investment through Messages' output_config.effort. Do not attempt to disable adaptive thinking for this model; complex tasks should be given sufficient max_tokens to avoid the final answer being truncated due to an insufficient budget.
Prefer Messages' native tool structure and tool_choice.type=auto. Do not use any or specify tool to force tool calls; omit temperature, top_p, and top_k. Tool execution is handled by the connected environment, and the model itself does not automatically gain access to external systems.
Billing is based on token usage, and different plans have different actual purchase costs.
Based on the current largest plan of 142,800 Credits／$13,584, standard input costs $3.20／million token and output costs $16／million token, which is 40% of the official standard API input and output prices (60% cheaper). Cache reads cost $0.08, 5-minute cache writes cost $4, and 1-hour cache writes cost $6.40／million token. For other plans, please see the pricing page; additional account discounts are subject to the actual invoice.
Messages reports standard input, cache reads, and cache writes separately; these components are billed separately. Chat Completions' total input includes cache details and must not be added again when calculating fees. Standard API price comparisons do not include special official pricing modes such as Batch and Fast.
Validate short tasks first, then gradually increase the material, tools, and task length.
For example: Read this code and error log, list the most likely causes, and provide a validation method for each cause. Please separate observations, inferences, and unverified content.
Choose the Messages API when you need native thinking content blocks and tool semantics; applications that already use the OpenAI message structure can use Chat Completions. Always specify the full model name claude-fable-5-2.
Check the text, finish reason, and usage. First confirm that short requests can complete, then adjust the output budget; for streaming calls, determine success only after reading the end event.
Five things to confirm before integration.
Use claude-fable-5-2. The older Fable 5.1 and Fable 5 remain separate models. When upgrading, explicitly change the model field and recheck results and costs.
Adaptive thinking remains enabled for this model. Omit temperature, top_p, and top_k, and control reasoning effort through the effort parameter supported by the selected API.
Messages retains Claude-native content blocks and tool interactions; Chat Completions is suitable for applications that already use the OpenAI message format. The fields and usage structures differ between the two formats and should be parsed according to their respective documentation.
The 60% comparison here uses the purchase cost of the current largest plan against the official standard API list price. Smaller plans, additional account discounts, and special pricing models must be calculated separately. Refer to the model pricing page and billing for specifics.
It is suitable for reasoning and planning multi-step tasks, but continued execution requires the application to manage tools, history, timeouts, and permissions. Split tasks into verifiable stages, and do not treat a single request as proof that an external operation has been completed.
Choose a development approach or application suited to this model and follow its setup guide.
OpenAI Chat Completions API
SDK installation, authentication and examples
Claude Code CLI Setup Guide
Coding tool setup · requires a Coding Plan application key
Using Ace Data Cloud with Cherry Studio
Chat client setup · Base URL, API key and model
Using Ace Data Cloud with ChatBox
Chat client setup · Base URL, API key and model