Claude Fable 5
Anthropic's first Mythos-class model for general use — state-of-the-art reasoning, coding, and agentic work with a 1M context window.
Overview
What is Claude Fable 5
Claude Fable 5 is Anthropic's most capable publicly available model, released June 2026. Built for demanding reasoning, software engineering, and long-horizon agentic tasks, it offers a 1M token context window, vision, tool use, and web search, with safety classifiers that occasionally reroute sensitive queries to Claude Opus 4.8.
Running it privately on Venice
On Venice, Claude Fable 5 runs under an anonymized privacy tier — your prompts are not stored, profiled, or used for training. You get the full frontier model with vision, reasoning, function calling, and web search, billed per token with no subscription lock-in. Note that Venice does not modify Anthropic's safety classifiers; the model remains the standard Fable 5 release.
Assessment
Strengths and limitations
- State-of-the-art performance on complex reasoning, software engineering, and scientific research benchmarks according to Anthropic's evaluations.
- 1M token context window and up to 128K tokens of output per request, enabling long-horizon agentic workflows over massive codebases or document sets.
- Multimodal: supports vision with multiple image inputs, function calling, reasoning, web search, code-optimized generation, and structured JSON output.
- Proactive and thorough on long-running, asynchronous tasks that previous generations could not sustain.
- Safety classifiers are designed to reduce misuse risks, with fallback to Claude Opus 4.8 when triggered.
- Closed and proprietary: no open weights, so self-hosting or fine-tuning is impossible.
- Safety classifiers can decline or silently reroute benign requests to Claude Opus 4.8; Anthropic notes this triggers on average in less than 5% of sessions, but it adds friction.
- Premium pricing: at $60 per 1M output tokens on Venice, it is among the most expensive options for high-volume generation.
- Not uncensored on Venice: retains Anthropic's built-in refusals and content policies.
- Does not run inside a TEE or with end-to-end encryption on Venice; privacy is handled via anonymization and zero retention instead.
Capabilities
What it supports
- Tool use / function calling
- Vision (image input)
- Reasoning
- Web search
- Code-optimized
- Structured output (JSON schema)
- Audio input
- Video input
- Multiple image inputs
- Log probabilities
Specifications
Datasheet
- Maker
- Anthropic
- Released
- June 9, 2026
- Model class
- Mythos-class (public release)
- Open weights
- No — proprietary
- Context window
- 1,000K tokens
- Max output
- 128K tokens
- Capabilities
- Vision, Function calling, Reasoning, Web search, Code-optimized
- Privacy on Venice
- Anonymized — prompts not stored
- Available on Venice since
- Jun 2026
API
Call it from your code
Venice exposes an OpenAI-compatible API. Point your base URL at Venice and pass the model id.
curl https://api.venice.ai/api/v1/chat/completions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-fable-5",
"messages": [{ "role": "user", "content": "Explain quantum tunneling simply." }]
}'Pricing
What it costs on Venice
Billed per token on Venice: $12 per 1M input tokens and $60 per 1M output tokens.
New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.
Alternatives
How it compares
| Model | Price (Venice) | Context window | Open weights | Best for |
|---|---|---|---|---|
| Claude Fable 5 | $12 in · $60 out / 1M | 1M tokens | No | Frontier reasoning & coding |
| Claude Opus 4.8 | $6 in · $30 out / 1M | 1M tokens | No | High-end Anthropic workloads |
| GPT-5.6 Sol | $6.25 in · $37.50 out / 1M | 1M tokens | No | OpenAI frontier tasks |
| DeepSeek V3.2 | $0.33 in · $0.48 out / 1M | 160K tokens | Yes | Open, cost-efficient reasoning |
| Kimi K2.6 | $0.75 in · $3.50 out / 1M | 256K tokens | Yes | Open-weight coding & analysis |
Anthropic's most capable public model with safety classifiers and 128K max output.
Use cases
What it is good for
- 01Agentic software engineering over large repositories and multi-file code generation.
- 02Long-form research synthesis and knowledge work across million-token document corpora.
- 03Multimodal analysis combining vision and text, such as interpreting charts, diagrams, and screenshots.
- 04Complex multi-step reasoning with live web search and tool use for autonomous task execution.
- 05High-stakes professional drafting where thoroughness and instruction-following matter more than latency.
Prompting
Getting better results
Build fallback logic: if Fable 5's classifiers trigger, the request may route to Opus 4.8. Verify responses when guaranteed Fable 5 capability is required.
Feed large contexts directly — the 1M window accepts entire codebases or book-length documents in one pass.
Combine vision with function calling to build agents that read UI screenshots and then call APIs.
Use structured output (JSON schema) for reliable parsing of long-form reasoning results.
Version history
Prior Anthropic flagship; now Fable 5's fallback.
Current — Mythos-class public release with safety classifiers.
Same weights as Fable 5, without classifiers; limited release.
FAQ
Frequently asked questions
Claude Fable 5 is Anthropic's first Mythos-class model released for general use, launched June 9, 2026. It is the company's most capable publicly available AI, built for demanding reasoning, software engineering, vision, and long-horizon agentic work, with a 1M token context window and safety classifiers that can reroute sensitive queries to Claude Opus 4.8.
On Venice, Claude Fable 5 costs $12 per 1M input tokens and $60 per 1M output tokens. Cached input is billed at $1.20 per 1M tokens. You pay per token with credits and no subscription is required.
It is not open source — the weights are proprietary and closed. On Venice you pay per token with credits to use it. There is no subscription required, and you only pay for what you generate.
Fable 5 outperforms Opus 4.8 on complex reasoning and coding, but it includes safety classifiers that may reroute some requests to Opus 4.8. Opus 4.8 is cheaper and avoids classifier friction, making it better for cost-sensitive or high-volume workflows where absolute frontier capability is not required.
Yes. On Venice, Claude Fable 5 supports function calling, vision with multiple image inputs, reasoning, web search, code-optimized generation, and structured JSON output.
They share the same underlying model. Claude Mythos 5 removes the safety classifiers and is available only to vetted Project Glasswing partners. Claude Fable 5 is the publicly accessible version with safeguards active.
Venice runs Claude Fable 5 under an anonymized privacy tier — your prompts are not stored, profiled, or used for training. However, it does not currently run inside a TEE or use end-to-end encryption.
Anthropic suspended access on June 12, 2026, shortly after launch, then restored it on July 1, 2026. The brief suspension was related to export control and safety review, which have since been resolved.
No. Venice provides an uncensored platform, but Claude Fable 5 itself retains Anthropic's safety classifiers and refusal behaviors. Venice does not modify the model weights to remove these safeguards.
Run Claude Fable 5 privately
No prompt logging. No data used for training.