Claude Fable 5.2 API pricing and discount
Call Claude Fable 5.2 through one OpenAI-compatible endpoint, billed $6.40 per 1M input tokens, 20% below Anthropic’s list price. Routes are verified before they serve you and re-checked continuously.
- 1M context
- Reasoning
- OpenAI-compatible
New accounts start with free credit. Then prepaid from $5, no subscription.
- Input, per 1M tokens
- $6.40was
$8.00 - Output, per 1M tokens
- $32.00was
$40.00
Same model, same tokens, less to pay. Billed per request from prepaid credits; every response reports the cost.
- 1Mtokens
- Context window
- 128ktokens
- Maximum output
- 20%off
- Below Anthropic’s list price
- 89%
- Of input usually cached
- 59%
- Of requests use reasoning
- Live
- Callable now, every key
What your Claude Fable 5.2 workload costs
Move the sliders to your request size and volume. It starts on a typical request from real traffic.
Starts on a typical Claude Fable 5.2 request (145.4k in, 1.7k out). Reasoning tokens bill as output; cached input bills at the normal input rate on OSMEXO today.
- Per request
- $0.987
- Per 1,000 requests
- $987
- You save / month
- $7,400
Estimate: (input × input price + output × output price) ÷ 1M per request, × requests per day × 30. Before any vendor cache discount.
Where Claude Fable 5.2 fits
What developers use Claude Fable 5.2 for, where it ranks highest, and what a typical request looks like, from OSMEXO Intelligence.
A typical request
145k tokens in · 1.7k out
- 89% of input tokens are usually cached
- Reasoning is on in 59% of requests
- 0.59 tool calls per request on average
Averages across recent traffic for Claude Fable 5.2; cache discounts not applied.
Call it from your code or coding agent
Change the base URL and the key; keep your code. Streaming, tool calls, structured output and vision behave as they do upstream, and every response carries the billed cost beside the list price.
anthropic/claude-fable-5.2from openai import OpenAI
client = OpenAI(
base_url="https://api.osmexo.com/v1", # was the vendor's URL
api_key="osm-...",
)
r = client.chat.completions.create(
model="anthropic/claude-fable-5.2",
messages=[{"role": "user", "content": "Hello"}],
)
Check it is the real Claude Fable 5.2, with your own key
Every route passes these checks against Anthropic’s own API before it serves anyone. Run the same battery yourself: one OpenRouter key verifies any text model, a direct Anthropic key works too, and exit code 0 means the same model.
- 1Tokenizer fingerprint
The same prompt must cost the same token count as on the vendor’s API.
- 2Knowledge cutoff probe
The route must name the same training cutoff and maker as the vendor’s model.
- 3Hidden-prompt probe
Asks for any operator system prompt, so a wrapper that injects one is caught.
- 4Exact-output tests
Deterministic tasks, such as reversing a string, must give the vendor’s answer.
- Approved route
Scored 0 to 100; 85 or more goes live, then is re-checked continuously.
curl -fsSLO https://dashboard.osmexo.com/verify.mjs
OSMEXO_API_KEY=osm-... VENDOR_API_KEY=sk-or-... \
node verify.mjs --model anthropic/claude-fable-5.2 --vendor openrouter
About Claude Fable 5.2 on OSMEXO
Price, fit, verification and failover. Anything else, ask us.
How much does Claude Fable 5.2 cost through OSMEXO?
$6.40 per 1M input tokens and $32.00 per 1M output tokens, billed per request from prepaid credits. That is 20% below Anthropic’s list price of $8.00 / $40.00. The price on this page refreshes from the catalog every few minutes.
Is this the same Claude Fable 5.2 as calling Anthropic directly?
Yes. Every route for Claude Fable 5.2 is probed against the vendor’s own API before it serves anyone: tokenizer fingerprint, knowledge cutoff, a hidden-prompt probe and exact-output tasks, scored 0 to 100 with 85 as the approval threshold. You can run the same check from your machine with your own vendor key; the command is on this page.
How do I call Claude Fable 5.2 with the OpenAI SDK?
Point the official OpenAI SDK at https://api.osmexo.com/v1 with your OSMEXO key and use the model id anthropic/claude-fable-5.2. Streaming, tool calls, structured output and vision work as they do upstream. Coding agents such as OpenCode, Cline, Roo Code, Continue and Aider take the same base URL and key.
What happens if a provider for Claude Fable 5.2 goes down?
The gateway retries and falls back to the next verified route for Claude Fable 5.2 before the first token is sent, so your users see a slower answer instead of an error. Every response reports which route answered, what it cost, and what the same tokens would have cost at list price.