Claude by Anthropic
Summary
Most frontier models hit a ceiling around the third chained tool call, where context degrades, instructions drift, and the agent starts hallucinating state it should have tracked — Claude Fable 5 is Anthropic's answer to that ceiling, built explicitly for long-horizon tasks where lesser models fall apart before the job is done.
Fable 5 runs on Anthropic's Mythos-class transformer architecture with adaptive thinking, giving it a 1M-token input context and up to 128k tokens of output — which means a codebase migration or a multi-document research synthesis fits in a single pass without chunking hacks. The vendor positions this explicitly for autonomous agent work: chained tool use, multi-step reasoning, and tasks where the model needs to hold complex state across many turns. Where it breaks is cost — per-token billing is paid-only, and at the rates the validator documents, teams running high-volume pipelines will feel it fast. Vision-dependent scientific analysis and complex software engineering are the use cases the vendor calls out directly. Teams doing commodity summarization or single-turn Q&A will pay a premium they cannot justify.
Bottom line: Bet on Fable 5 when you need an agent to hold a codebase in memory and reason across it without losing the thread — switch to a cheaper model when your workload is high-volume and the task is simple enough that quality differences disappear in your eval.
Pricing Plans
Per-token- Price
- $10 per million input tokens, $50 per million output tokens
- Free Tier
- No free tier. Limited free trial via subscription plans through June 22, 2026.
API Standard
Pay-as-you-go API access at standard rates
- $10 per M input tokens
- $50 per M output tokens
- 1M token context window
- 128k max output tokens
- Full feature access
Subscription (through June 22, 2026)
Included in Pro ($20/mo), Max ($100/mo), Team, and Enterprise plans through June 22, 2026 only
- No additional charge through June 22
- Usage credits required after June 23
- Available on web, desktop, iOS, Android
View full pricing on claude.ai →
Pricing may have changed since last verified. Check the official site for current plans.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- 1M-token input context, so a full codebase or multi-document corpus fits in one pass without chunking pipelines that introduce retrieval errors and context fragmentation.
- Up to 128k output tokens per response, which means the model can return a complete migration script or exhaustive technical analysis in a single call rather than forcing you to stitch together multiple truncated completions.
- Adaptive thinking architecture, per vendor documentation, adjusts reasoning depth to task complexity — so multi-step agent tasks that cause shallower models to drift or lose state have a higher ceiling before requiring human correction.
- Native tool use with multi-step chaining, so agents can plan, call external tools, evaluate results, and continue reasoning without you writing glue logic to re-inject context between steps.
- Provider-direct API with Anthropic's Constitutional AI alignment focus, which means safety-critical applications get a model that is less likely to produce confidently wrong or harmful outputs mid-agent-run — reducing the failure modes that are hardest to catch in automated pipelines.
Cons
Sign in to edit- Per-token billing at the rates the validator documents makes high-volume pipelines expensive fast — teams running thousands of structurally similar, low-complexity requests will find that cost per useful output is worse than lighter models, and the standard path is to route those workloads to GPT-5.5, Gemini 3.1 Pro, or a self-hosted Llama 4 deployment depending on latency and privacy needs.
- No self-hosted option exists — full stop — so teams with data residency requirements, air-gapped infrastructure, or procurement rules that prohibit third-party API calls for sensitive data cannot deploy this model regardless of quality, and the competitor they move to is whatever open-weight model fits their compliance posture.
- Long-context performance at the upper end of the 1M-token window is a vendor claim the scraped source page does not corroborate with third-party benchmarks — teams building pipelines that depend on reliable recall at 800k+ tokens should validate this against their own workload before committing architecture decisions to it.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- Claude API, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry, GitHub Copilot, Claude Code, Claude Platform on AWS, claude.ai
- API Available
- Yes
- Self-Hosted
- No
- Last Updated
- 2026-06-11T05:03:36.579Z
Best For
Who it's for
- Frontier AI research and development
- High-complexity agentic coding workflows
- Vision-heavy applications requiring state-of-the-art accuracy
- Long-context reasoning on specialized domains
- Tasks where solution quality justifies 2x API costs
What it does well
- Complex software engineering and codebase migrations
- Long-horizon autonomous agent orchestration
- Vision-dependent scientific research and analysis
- Knowledge-intensive writing and reasoning tasks
- Multi-step problem solving with tool integration
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Similar Tools
Compare Claude by Anthropic
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is Claude by Anthropic free?
- Claude by Anthropic is a paid tool ($10 per million input tokens, $50 per million output tokens). No permanent free tier is offered.
- Is Claude by Anthropic open source?
- No — Claude by Anthropic is a closed-source tool. Source code is not publicly available.
- Does Claude by Anthropic have an API?
- Yes. Claude by Anthropic exposes a developer API. See the official documentation at https://claude.ai for details.
- When was Claude by Anthropic released?
- Claude by Anthropic was first released in 2026.
- What platforms does Claude by Anthropic support?
- Claude by Anthropic is available on: Claude API, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry, GitHub Copilot, Claude Code, Claude Platform on AWS, claude.ai.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
Claude Fable 5 is a closed, API-accessed frontier model from Anthropic, designed for tasks that demand sustained reasoning across long contexts and multi-step tool use. The core workflow is API-driven: you send structured prompts with tool definitions, the model reasons through a problem using adaptive thinking, calls tools, processes results, and continues — holding up to 1M tokens of input context across the entire exchange. Output reaches up to 128k tokens per response, which matters when the deliverable is a full migration plan, a detailed analysis, or a long-form technical document rather than a short completion.
The differentiating architecture claim is adaptive thinking — a vendor-described capability where the model adjusts its internal reasoning depth based on task complexity. In practice, the vendor markets this as what separates Fable 5 from its own prior models for long-horizon agentic work: the model doesn’t just respond, it plans, backtracks, and re-evaluates before committing output. For autonomous agent orchestration — where the model is coordinating tool calls across many steps without a human reviewing each one — this matters more than raw benchmark scores.
Fable 5 fits teams where solution quality is the primary constraint and cost is secondary: frontier AI research, vision-heavy scientific pipelines, high-complexity codebase work, and specialized domain reasoning where cheaper models produce errors that cost more to fix than the API bill. It does not fit high-volume, low-complexity pipelines. Teams hitting per-token costs at scale — especially those running thousands of short, structurally similar requests — report switching to lighter models or self-hosted alternatives like Meta Llama 4, which offers a cost ceiling Fable 5 cannot match by design.
Access is exclusively through Anthropic’s API; there is no self-hosted option. This is a hard constraint for teams with data residency requirements or air-gapped deployment environments — no workaround exists within this product.