Long-Horizon Agentic Work
Handles multi-stage tasks that run for hours, use multiple tools, recover from failed steps, and continue toward a defined result with less supervision.
Claude Fable 5.1
Access Anthropic’s highest-capability generally available model for long-running coding, research, and multi-stage knowledge work through one GPTProto API key and a shared balance across 200+ AI models.
Claude Fable 5.1 pricing
Estimate a request with real work scenarios. GPTProto token pricing is 10% below official rates.
用量數量單價小計
tokens
tokens
tokens
tokens
單次請求費用$0.0888
請求次數
儲值金額
儲值 $100 你將獲得：
儲值額度永久有效。你總共會收到 $100.00。
額外 10% 模型折扣，相較官方 Claude API 可節省 $11.0937。
相關模型
所有模型
| 模型 | 輸入 → 輸出 | ||||
|---|---|---|---|---|---|
| — | $9.00 / $45.00 每 1M | $11.25 / $0.23 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | |
| — | $1.80 / $5.40 每 1M | $2.25 / $0.23 每 1M | 輸入: 文字輸入: 圖像輸入: 影片輸入: 文件輸入: 音訊 輸出: 文字 | ||
| — | 1.31M | $0.15 / $0.50 每 1M | — / $0.03 每 1M | 輸入: 文字輸入: 圖像輸入: 影片輸入: 文件 輸出: 文字 | |
| — | 1.05M | $0.44 / $1.32 每 1M | — / $0.01 每 1M | 輸入: 文字輸入: 圖像 輸出: 文字 | |
| 1.31M | $1.26 / $3.96 每 1M | — / $0.23 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $0.90 / $4.50 每 1M | $0.60 / $0.09 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 500K | $1.20 / $3.60 每 1M | — / $0.30 每 1M | 輸入: 文字輸入: 圖像 輸出: 文字 | ||
| 1M | $1.80 / $5.40 每 1M | $2.25 / $0.23 每 1M | 輸入: 文字輸入: 圖像輸入: 影片輸入: 文件 輸出: 文字 | ||
| 1M | $4.50 / $22.50 每 1M | $5.63 / $0.45 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $0.90 / $4.50 每 1M | $0.60 / $0.09 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $0.18 / $1.50 每 1M | $0.60 / $0.02 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $2.70 / $13.50 每 1M | $0.27 / $0.27 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $0.16 / $0.96 每 1M | $0.20 / $0.02 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $1.60 / $9.60 每 1M | $2.00 / $0.16 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $3.20 / $16.00 每 1M | $4.00 / $0.32 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 500K | $1.20 / $3.60 每 1M | $0.30 / $0.30 每 1M | 輸入: 文字輸入: 圖像 輸出: 文字 | ||
| 1M | $1.80 / $9.00 每 1M | $2.25 / $0.18 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 1.05M | $0.48 / $0.96 每 1M | $0.10 / $0.10 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1.05M | $1.26 / $3.96 每 1M | $0.23 / $0.23 每 1M | 輸入: 文字輸入: 圖像輸入: 文件 輸出: 文字 | ||
| 1M | $9.00 / $45.00 每 1M | $11.25 / $0.90 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 1M | $0.36 / $1.44 每 1M | $0.07 / $0.07 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 1M | $4.50 / $22.50 每 1M | $5.63 / $0.45 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 1M | $4.50 / $22.50 每 1M | $5.63 / $0.45 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| — | 1.05M | $0.44 / $1.32 每 1M | — / $0.01 每 1M | 輸入: 文字 輸出: 文字 | |
| — | 1.05M | $1.32 / $3.96 每 1M | — / $0.04 每 1M | 輸入: 文字 輸出: 文字 | |
| 1M | $0.75 / $1.50 每 1M | $0.12 / $0.12 每 1M | 輸入: 文字輸入: 圖像 輸出: 文字 | ||
| 262K | $0.85 / $3.60 每 1M | $0.14 / $0.14 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 1M | $4.50 / $22.50 每 1M | $5.63 / $0.45 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 205K | $0.24 / $0.96 每 1M | $0.30 / $0.02 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 262K | $0.54 / $2.70 每 1M | $0.09 / $0.09 每 1M | 輸入: 文字輸入: 文件 輸出: 文字 | ||
| 262K | $0.10 / $0.97 每 1M | — | 輸入: 文字輸入: 圖像 輸出: 文字 | ||
| 262K | $0.10 / $0.97 每 1M | — | 輸入: 文字輸入: 圖像 輸出: 文字 | ||
| 262K | $0.02 / $0.18 每 1M | — | 輸入: 文字輸入: 圖像 輸出: 文字 |
Claude Fable 5.1 API
Build coding agents, research systems, and complex document workflows with Claude Fable 5.1 API. It combines a 1M-token context window, up to 128K output tokens, and always-on adaptive thinking. Access it on GPTProto at 10% below Anthropic’s standard input and output rates.
Codebase-Level Engineering
Supports repository-wide features, code review, performance analysis, test creation, root-cause debugging, and implementation across multiple files.
1M Context and 128K Output
Process large repositories and technical documentation within a 1M-token context window, then return up to 128K tokens in one response.
Always-On Adaptive Thinking
Reasoning stays enabled. Tune effort from low to max to balance task difficulty, latency, and token use instead of switching thinking off.
What Is Claude Fable 5.1?
Claude Fable 5.1 is Anthropic’s most capable generally available model for demanding reasoning and long-horizon agentic work. Released on September 1, 2026, it succeeds Claude Fable 5 while retaining the same 1M-token context window, 128K maximum output, and base list price. It is a proprietary hosted model designed for difficult work that must maintain quality across many steps.
| Specification | Claude Fable 5.1 |
|---|---|
| Release date | September 1, 2026 |
| Official API model ID | claude-fable-5-1 |
| Context window | 1,000,000 tokens |
| Maximum output | 128,000 tokens |
| Thinking | Adaptive, always on |
| Effort | High by default; low to max supported |
| Knowledge cutoff | June 2026 |
Claude Fable 5.1 Upgrades That Matter in API Workloads
Fable 5.1 improves most clearly on work combining reasoning, tools, and long execution chains. Anthropic reports 55.8% on Terminal-Bench 4.0 versus 42.0% for Fable 5. Its direct cache-read price also fell from $1 to $0.25 per million tokens, with Anthropic estimating about 25% lower typical workload cost and savings of up to approximately 45% for highly agentic work. Actual savings depend on cache reuse.
Claude Fable 5.1 Applications
| Application | Why Fable 5.1 fits |
|---|---|
| Agentic coding | Plans repository changes, writes tests, checks results, and recovers from failed steps. |
| Long-running agents | Maintains direction across extended tool sequences and asynchronous work. |
| Research | Synthesizes large source collections and follows multi-part instructions. |
| Knowledge work | Produces document, spreadsheet, presentation, and operational deliverables for review. |
Claude Fable 5.1 vs Fable 5, Opus 5, and GPT-5.6
Anthropic’s published benchmark set indicates that Fable 5.1 is intended for tasks where long-horizon completion quality matters more than minimum latency or token price.
| Anthropic-published evaluation | Fable 5.1 | Fable 5 | Opus 5 | GPT-5.6 Sol |
|---|---|---|---|---|
| Terminal-Bench 4.0 | 55.8% | 42.0% | 52.3% | 37.3% |
| AutomationBench | 31.4% | 17.1% | 26.9% | 19.6% |
| CursorBench 3.2.0 | 73.4% | 70.5% | 70.0% | 67.2% |
These are provider-run results, not independent rankings. For GPT-5.6, Grok 4.6, Qwen3.8-Max-0902, Kimi K3, or GLM-5.3 Flash comparisons, use identical tools, context, time limits, and acceptance tests. Choose Fable 5.1 when failed completion across a long workflow is more expensive than its higher token price.
What Developers Should Check Before Migrating
Moving from Claude Fable 5 or Opus 5 requires more than changing the model name:
Forced tool choice is unsupported.
tool_choicevaluesanyandtoolreturn a 400 error. Useautowith explicit instructions and strict tool schemas, or request structured JSON.Thinking cannot be disabled. Control spend and latency through
effort; test medium for routine stages and reserve xhigh or max for evaluation-proven gains.Keep conversation history append-only. Editing earlier messages can invalidate preserved thinking blocks. After client-side compaction, remove stale thinking blocks or restart from a clean summary.
When Should You Use Claude Fable 5.1 API?
Use Claude Fable 5.1 for work that is difficult, long-running, and expensive to restart: repository-scale engineering, open-ended debugging, multi-document research, or agents coordinating several tools. Use a lower-cost model for routine chat, extraction, classification, and short code generation.
FAQ
How much does Claude Fable 5.1 API cost on GPTProto?
GPTProto pricing is $9 per million input tokens and $45 per million output tokens, 10% below Anthropic’s $10/$50 list rates. Cached input and batch rates may follow separate rules.
How can I get a Claude Fable 5.1 API key?
Create a GPTProto account and generate a key from the dashboard. The same key and shared balance work across Claude Fable 5.1 and 200+ other models.
Is Claude Fable 5.1 open source?
No. It is a proprietary Anthropic model available through hosted API services; its weights and training data are not released for local deployment.
What is the difference between Claude Fable 5.1 and Claude Fable 5?
Both have a 1M context window, 128K output, and the same Anthropic base price. Fable 5.1 improves long-running work, cuts direct cache-read pricing by 75%, and changes tool-choice and preserved-thinking behavior.
Is Claude Fable 5.1 better than GPT-5.6?
Anthropic reports higher Fable 5.1 scores than GPT-5.6 Sol on three published agentic evaluations. Because these are provider-run results, compare both APIs on your own tools, latency target, budget, and acceptance tests.
Can I test Claude Fable 5.1 in an API playground?
Yes. Use the GPTProto playground before integration. Test representative long tasks rather than short chat prompts, and compare cost per accepted result—not only token price—when choosing models or effort levels.
相關文章
與本模型相關的指南、對比與更新。
所有文章
Introducing Claude Fable 5.1 and Claude Mythos 5.1: Same Model, Different Safeguards
Claude Fable 5.1 and Mythos 5.1 share one model but differ in access. See pricing, features, benchmarks, safeguards, and API migration changes.
GLM 5.3 Flash vs DeepSeek V4 Flash: Which Is Better for Code, Agents, and Cost?
Meta Description: Compare GLM 5.3 Flash vs DeepSeek V4 Flash for coding, frontend work, agents, speed, context, and API pricing to choose the better model.
Qwen3.8-Flash-Next vs GLM-5.3 Flash: Which Is Better for Coding, Agents, and Price?
Qwen3.8-Flash-Next vs GLM-5.3 Flash compared for coding, agents, frontend work, speed, pricing, context, licenses, and production use.
What Is Hunyuan 4? Tencent Hy4 Preview Features, Pricing, Benchmarks, and Release Status
What is Tencent Hy4 Preview? See its release status, 770B MoE design, 1M context, API pricing, benchmarks, limits, and model comparisons.