Blank white background with no objects or features visible.

Hello, EMEA! TrueFoundry is growing our footprint across Europe, the Middle East, and Africa. Read more. →

Claude Fable 5 vs Opus 4.8: Benchmarks, Pricing and What Has Changed Since

By Deepti Shukla

Updated: November 16, 2026

Published: June 10, 2026

Fable 5 vs pOpus 4.8

‍

⚡ TL;DR
  • Fable 5.2 ($8 in / $40 out per million tokens) is the top tier. Opus 5.5 ($4 / $20) is the cheaper step down for most work.
  • The original Fable 5 vs Opus 4.8 gap was real: 80.3% vs 69.2% on SWE-Bench Pro. Both models have since been replaced by newer versions.
  • Pick by task, not by brand. Route hard, long-running work to Fable and everything else to Opus.

When Claude Fable 5 launched on June 9, 2026, it gave developers a second tier above Opus for the first time. The first question was simple: is the jump worth double the price? This post answers that for Fable 5 and Opus 4.8, the pair most people searched for, and then brings the picture up to date.

Short version: Fable 5 beat Opus 4.8 by a wide margin on hard coding and long-running work. Since then Anthropic has shipped Opus 5, Opus 5.5, Fable 5.1 and Fable 5.2, and both original models have been replaced. The gap and the price question haven't gone away, they've just moved up a version.

Where the lineup stands today

As of November 16, 2026, these are the models in play. Prices are from Anthropic's pricing page, and all of them have a 1M-token context window and 128K output.

ModelReleasedInput / output (per 1M tokens)Where it fits
Fable 5.2Nov 9, 2026$8 / $40Hardest coding and long-running agent work
Fable 5.1Autumn 2026$10 / $50Superseded by 5.2
Fable 5Jun 9, 2026$10 / $50The model this post was first written about
Opus 5.5Sep 22, 2026$4 / $20Strong default for most production work
Opus 5Jul 24, 2026$5 / $25Superseded by 5.5
Opus 4.82026$5 / $25The original comparison point

How do Fable 5 and Opus 4.8 compare?

Fable 5 was built for demanding reasoning, software engineering, research and long-running tasks. Opus 4.8 was aimed at agentic coding and everyday enterprise work. Here's the launch-day comparison:

Claude Fable 5Claude Opus 4.8
ClassMythos-class, top tierOpus class
SWE-Bench Pro80.3%69.2%
FrontierCode29.3%13.4%
Input price$10 per 1M tokens$5 per 1M tokens
Output price$50 per 1M tokens$25 per 1M tokens
Context window1,000,000 tokens1,000,000 tokens
Max output128K tokens128K tokens
Model stringclaude-fable-5claude-opus-4-8

Fable 5 scored 11.1 points higher on SWE-Bench Pro and 15.9 points higher on FrontierCode. The gap shows up most on large codebases, multi-file debugging and tasks that run for hours. For a small bug fix, you probably won't notice it.

Benchmarks measure specific tasks under specific conditions, so treat them as a starting point. Run your own prompts and track completion rate, retries, latency, token use and how often a human has to step in. Our guide to benchmark scores in production explains why.

How much does each cost?

At launch Fable 5 cost twice as much as Opus 4.8 on every line:

InputOutput5-min cache write1-hour cache writeCache hit
Fable 5 (launch)$10$50$12.50$20$1
Opus 4.8$5$25$6.25$10$0.50

The sticker price overstates the gap a little. Fable often finishes in fewer turns, and you don't pay Fable rates for requests that get rerouted to Opus. Even so, for high-volume or simple work, the cheaper model usually wins. Newer pricing narrowed the spread in absolute terms: Fable 5.2 is $8 / $40, and Opus 5.5 is $4 / $20. The ratio is still 2 to 1. For more on keeping spend down, see our LLM cost optimization guide.

Not sure which Claude model fits which job?

Send a slice of your real traffic to both and compare cost, latency and quality side by side. TrueFoundry's AI Gateway puts Fable and Opus behind one endpoint, so the test takes minutes.

See it on your workloads → Book a demo

Is Fable slower than Opus?

Usually, yes, on hard tasks. Fable works through more steps before it answers, which is where the quality comes from and also where the extra seconds go. On short prompts the difference is small. If your app is latency-sensitive, such as a chat UI or an autocomplete, Opus is the safer pick. If a job runs in the background, the wait rarely matters.

What actually differs between them?

Reasoning and coding. Fable holds up better on repository-level changes and problems that need many dependent steps. Opus is plenty for scoped tasks.

Long-running work. Fable stays on track across long agent runs more reliably, so it needs fewer restarts and less hand-holding.

Limits. Both have a 1M-token context window and 128K output, so context size won't decide it.

Safeguards. Fable includes classifiers for cybersecurity, biology and chemistry, and distillation. Anthropic says fewer than 5% of sessions trigger them.

What happens when Fable falls back to Opus?

If a Fable request trips a safeguard, it can be handed to Opus, and the user is told. In the Claude apps this happens automatically. API customers set it up through Anthropic's Fallback API, so it isn't on by default. Plan for it: treat the two models as a pair and make sure your app handles an answer that came from the other one.

When to use each

Choose Fable for long multi-stage work such as large migrations or multi-day agent runs, for hard problems where quality matters more than token cost, and for deep research or coding where Opus has stopped improving.

Choose Opus for well-defined routine tasks, anything where latency or cost comes first, and high-volume traffic where double pricing adds up fast.

The better pattern: route between both

Most teams don't need to pick one. Send each request to the model that fits it. Useful signals are task complexity, traffic volume, per-team budget and what should happen on an error or rate limit. Default to the cheaper model and escalate when the task needs it.

An AI gateway makes this practical because both models sit behind one OpenAI-compatible endpoint. Changing models is a one-line change, not a new integration.

Put a budget on every model your teams use

Set per-team spend limits, rate limits and fallbacks across Anthropic, Bedrock and Vertex, and see the cost of each request.

Book a 20-min demo →Read the AI Gateway docs

How to set up routing with TrueFoundry

Step 1. In your connected Anthropic provider, open Models Selection and enable the Fable and Opus models you want. Use Access Control to decide which teams and apps can call each one.

Claude Fable available in the TrueFoundry AI Gateway model list

Step 2. Open the Playground, pick a model and copy the usage snippet. To switch models later, change the model ID.

Switching between Claude models through the same TrueFoundry gateway

Once routing is live, watch cost per request, latency, completion rate, retries, errors and how often fallback fires. Those numbers tell you whether the split is working. Our AI usage analysis shows how to break them down by team. For the full model-by-model rundown, see Claude Fable 5 API, benchmarks and pricing.

For platform teams

Route every request to the cheapest model that can do the job

Model routing, budgets and failover for Claude and every other provider, running in your own cloud.

Try now.

One gateway for all your models, MCP servers, and agents. No credit card needed.

Start free
Table of Contents

One Gateway for Every LLM, Agent and MCP Server

Book a 30-min with our AI expert

Book a Demo

The fastest way to build, govern and scale your AI

Book Demo
Summarize with
ChatGPT logo by OpenAI
Perplexity AI logo
Blurry red snowflake on white background, symmetrical frosty design with soft edges and abstract shape.

Discover More

November 5, 2025
|
5 min read

Data Residency in the Age of Agentic AI: How AI Gateways Enable Sovereign Scale and Compliance

August 27, 2025
|
5 min read

Mapping the On-Prem AI Market: From Chips to Control Planes

August 27, 2025
|
5 min read

AI Gateways: From Outage Panic to Enterprise Backbone

Secure AI Gateway with MCP: Enterprise-Ready Protection
July 4, 2025
|
5 min read

Secure AI Gateway with Centralized MCP for Enterprises

November 18, 2026
|
5 min read

Claude Code Limits Explained: Pro, Max and Ultra

No items found.
Analyzing how Cloudflare detects MCP traffic
November 18, 2026
|
5 min read

How Cloudflare Detects MCP Traffic and Helps Secure It

No items found.
November 18, 2026
|
5 min read

What Is AI Search? How It Works, Architecture, and Enterprise Use Cases

No items found.
Explaining LLM fine-tuning for enterprise AI teams
November 18, 2026
|
5 min read

A Practical Guide to LLM Fine-Tuning: Methods, Data, Costs, and Deployment

No items found.
No items found.

Recent Blogs

Black left pointing arrow symbol on white background, directional indicator.
Black left pointing arrow symbol on white background, directional indicator.

Frequently asked questions

Is Claude Fable 5 better than Opus 4.8?

Anthropic’s published benchmarks show Fable 5 ahead on several demanding coding and reasoning tests, including SWE-Bench Pro and FrontierCode. The difference is more relevant for complex, long-running tasks; for routine workloads, Opus 4.8 may still meet your requirements at a lower cost.

‍

How much more expensive is Fable 5?

Fable 5 costs $10 per million input tokens and $50 per million output tokens. Opus 4.8 costs $5 and $25 respectively. That makes Fable 5 roughly twice the price on both input and output tokens.

‍

Why would a Fable 5 request return an Opus 4.8 answer?

Anthropic says some flagged cyber, bio-chemical, or model-distillation requests may be routed from Fable 5 to Opus 4.8. This applies to a small share of sessions, and those requests are billed at the Opus 4.8 rate rather than the Fable 5 rate.

‍

What can Fable 5 do that Opus 4.8 can't?

Fable 5 is positioned for more demanding reasoning, coding, and long-horizon agentic workloads. However, benchmark differences do not mean that Opus 4.8 cannot perform these tasks. The practical difference depends on task complexity, accuracy requirements, cost, and how much supervision your workflow needs.

‍

What is the difference between Fable 5 and Opus 4.8?

Fable 5 is the newer, higher-priced model, while Opus 4.8 costs about half as much per input and output token. Their reported benchmark performance also differs, particularly on complex coding and reasoning workloads.

‍

Can I use Fable 5 and Opus 4.8 through the same integration?

Yes. Both models can be accessed through the same TrueFoundry AI Gateway endpoint. You can route requests to either model using its model ID, allowing your application to switch between models without maintaining separate integrations.

‍

Do Fable 5 and Opus 4.8 have the same context window?

Check the current model documentation before implementation, as context limits can vary by model and API configuration. Context window size should be evaluated alongside output limits, pricing, and actual prompt requirements rather than treated as the only capacity metric.

‍

Should I compare Fable 5 with newer Claude models?

Yes. If you are selecting a model for a new deployment, compare the models currently available for your platform rather than relying only on a Fable 5 vs Opus 4.8 comparison. Anthropic's current documentation lists newer models, including Fable 5.1 and Opus 5.5.

‍

Is Opus 4.8 still worth using instead of Fable 5?

For workloads where Opus 4.8 provides sufficient quality, its lower token price can make it a practical option. Teams can also use both models, routing more demanding tasks to Fable 5 while keeping routine or high-volume traffic on Opus 4.8.

‍

Take a quick product tour
Start Product Tour
Product Tour