claude-fable-5Claude Fable 5 is Anthropic's flagship Mythos-class model optimized for long-horizon reasoning, agentic coding, and autonomous workflows. It supports text and image inputs with text output, featuring a 1M-token context window and 128K maximum output. Di Hostlinku: Performance mulai 2.5/1M input · Low-Cost mulai 1.77/1M input. Harga Original $10/1M input.
Hostlinku price · per 1M token
PerformancePerformance · Hemat hingga 50% · Kualitas & stabilitas terbaik
Input
/ 1M token
Output
/ 1M token
Cache read
per 1M token
0.2495
$1Cache write (5m)
per 1M token
3.12
$12.50Cache write (1h)
per 1M token
4.99
$20Web search
per 1K calls
2.5
$10Billing dalam (1 = $1). Coret = harga resmi vendor (hanya bila harga lebih murah). Service Mode dipilih per API key.
Dapatkan API Key + freeContext window
1.0M
1,000,000 tokens
Max output
128.0K
128,000 tokens
7-day uptime
99.60%
252 probes · 1 failed
Measured throughput
16.9 tok/s
Real traffic, last 7 days · output side · 12.86s avg
What this model can do through the Hostlinku gateway.
Vision
Reads images, screenshots, and diagrams alongside text. Send image inputs directly in the messages payload.
Function calling
The model calls tools you define and folds their results into the reply. Required for agentic workflows and multi-step automation.
Reasoning
Spends extra thinking tokens on hard problems before answering. Best for math, code, and multi-step logic.
Prompt caching
Response schema
Guarantees JSON output that matches a schema you provide. No brittle post-parsing on your side.
Streaming
Streams tokens as they are generated for real-time UX. Supported on every Hostlinku protocol.
Web search
Runs server-side web searches and folds the results into the answer. Billed per 1K calls.
Audio
Video
File input
Accepts documents such as PDFs and spreadsheets as context. Great for Q&A over long attachments.
Realtime
Session
Media
Multipart form
Binary response
Hostlinku probes this model on a schedule: measured, not self-reported.
99.60%
252 probes in 7 days, 1 failed · automatic failover to backup channels
Latency diukur dari probe terjadwal melalui supply chain nyata; throughput dari traffic user 7 hari terakhir (sisi output).
Swap the base URL and everything else stays the same. Native protocols pass through, adapted protocols are translated by the Hostlinku gateway.
curl https://api.hostlinku.com/v1/messages \
-H "x-api-key: $HOSTLINKU_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "content-type: application/json" \
-d '{
"model": "claude-fable-5",
"max_tokens": 1024,
"messages": [
{ "role": "user", "content": "Explain edge failover in one paragraph." }
]
}'Setiap dimensi yang ditagih, dengan harga resmi vendor (dicoret bila lebih mahal) di sebelah harga per Service Mode.
| Item | Unit | Original ($) | Performance () | Low-Cost () | Dedicated () |
|---|---|---|---|---|---|
| Input (prompt) | per 1M tokens | $10 | 2.5 | 1.77 | — |
| Output (completion) | per 1M tokens | $50 | 12.48 | 8.85 | — |
| Cache read | per 1M tokens | $1 | 0.2495 | 0.177 | — |
| Cache write (5m) | per 1M tokens | $12.50 | 3.12 | 2.21 | — |
| Cache write (1h) | per 1M tokens | $20 | 4.99 | 3.54 | — |
| Web search | per 1K calls | $10 | 2.5 | 11.8 | — |
Harga resmi vendor dicoret bila setidaknya satu Service Mode lebih murah. Dimensi tanpa coretan (mis. web search) tidak didiskon upstream. 1 = $1.
Model lain dari vendor yang sama, dengan harga dan konteks pembanding.
Pertanyaan umum soal Claude Fable 5: harga, kapabilitas, dan cara mengukurnya.
Anthropic
Satu API key untuk semua model. Ganti base URL, pakai SDK yang sudah kamu gunakan sekarang, dan bayar per token dalam (1 = $1). Pilih Service Mode Performance atau Low-Cost per API key.