HomeArtificial IntelligenceClaude Models 2026: Fable 5.2 vs Opus 5.5 vs Sonnet 5.5

Claude Models 2026: Fable 5.2 vs Opus 5.5 vs Sonnet 5.5

Last updated: November 12, 2026 · Author: Ignacy Kwiecień, founder & editor-in-chief, DecodeTheFuture.org

Updated November 12, 2026

Claude Fable 5.2, released on November 9, 2026, replaces Fable 5.1 in Anthropic’s current lineup at $8 input / $40 output per million tokens, down from $10 / $50. Fable 5.1 moved to the legacy section and stays active; Claude Mythos 5.2 replaces Mythos 5.1 for verified organizations. Prices, IDs, latency labels and retirement dates were rechecked against Anthropic’s documentation on this date.

Anthropic’s current Claude lineup has four models: Claude Opus 5.5 is Anthropic’s recommended starting point for most workloads ($4 input / $20 output per million tokens); Claude Fable 5.2, released November 9, 2026, is for demanding reasoning and long-horizon agentic work ($8 / $40); Claude Sonnet 5.5 is the best combination of speed and intelligence ($2 / $10); and Claude Haiku 5.5 is the fastest and cheapest, from $0.10 / $0.50 for prompts up to 100,000 tokens. All four have a 1M-token context window, 128K maximum output, adaptive thinking and a June 2026 knowledge cutoff.

Claude Opus 5.5 Claude Fable 5.2 Claude Sonnet 5.5 Claude Haiku 5.5 Official API pricing

Current Claude models at a glance

Anthropic’s official models overview lists Fable 5.2, Opus 5.5, Sonnet 5.5 and Haiku 5.5 as the current comparison set. All four accept text and image input, return text, support vision, tool use and multilingual work, and are available on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. Each has a 1M-token context window and up to 128K output tokens. Their roles, prices and thinking defaults differ.

ModelAPI model IDOfficial positioningLatencyReleasedInput / output per 1M tokens
Claude Opus 5.5claude-opus-5-5Long-running agentic coding and knowledge work; Anthropic’s default starting pointModerateSept. 22, 2026$4 / $20
Claude Fable 5.2claude-fable-5-2Demanding reasoning and long-horizon agentic workModerateNov. 9, 2026$8 / $40
Claude Sonnet 5.5claude-sonnet-5-5Best combination of speed and intelligenceFastSept. 28, 2026$2 / $10
Claude Haiku 5.5claude-haiku-5-5High-volume, latency-sensitive tasks such as classification, extraction and routingFastestOct. 7, 2026$0.10 / $0.50 (prompts up to 100K tokens); $0.50 / $2.50 above
Haiku 5.5 pricing has a prompt-size boundary.

The $0.10 input / $0.50 output rates apply to prompts up to 100,000 tokens. For prompts over 100,000 tokens, Haiku 5.5 costs $0.50 input and $2.50 output, and cache reads rise from $0.01 to $0.05. Prices are per million tokens before batch (50% off) or prompt-caching discounts.

Which Claude model should you use?

Anthropic’s own guidance is now explicit: start with Opus 5.5 for most workloads, and move to Fable 5.2 for demanding reasoning and long-horizon agentic work, or when your evals on Opus 5.5 at higher effort still fall short. For cost-sensitive production, test Sonnet 5.5 and Haiku 5.5 against the same evaluation set and move down a tier whenever quality holds. Anthropic recommends building that set around real prompts, edge cases, accuracy, latency and cost.

When you need…Start withWhy
A default for agentic coding, knowledge work and most workloadsOpus 5.5Anthropic’s recommended starting point; built for multihour coding agents, large refactors and computer use.
The hardest reasoning and long-horizon agentsFable 5.2Anthropic’s highest tier for demanding reasoning, at 2x Opus 5.5’s token price.
Everyday coding, agents and enterprise workflows at lower costSonnet 5.5Half Opus 5.5’s per-token price with fast responses.
Classification, extraction, routing and sub-agents at high volumeHaiku 5.5Lowest price and fastest latency in the current lineup, now with a 1M-token context window.

For coding-tool context, compare Cursor and Claude Code. For task-specific validation, use a fixed benchmark suite rather than vendor positioning alone; our AI agent benchmarks guide explains how to structure those tests. If you use Claude through a subscription rather than the API, plan limits matter more than token prices: see the Claude Ultra plan and current usage limits.

Current Claude models by relative capability and API input price A qualitative chart places Haiku 5.5 at the lowest price, Sonnet 5.5 above it, Opus 5.5 higher, and Fable 5.2 at the highest capability and price. Haiku 5.5 shows its input price for prompts up to 100,000 tokens. Claude model capability and price mapQualitative comparison based on Anthropic’s current model positioning and API input prices, checked November 12, 2026. API input price per 1M tokens → Relative capability → Haiku 5.5 from $0.10 input Sonnet 5.5 $2 input Opus 5.5 · default $4 input Fable 5.2 $8 input

The vertical scale is qualitative, reflecting Anthropic’s positioning rather than a cross-domain benchmark score. Input prices are official API rates; Haiku 5.5 shows its rate for prompts up to 100,000 tokens.

Claude Opus 5.5: the default starting point

Claude Opus 5.5 is the current Opus model, released on September 22, 2026, and the model Anthropic tells developers to start with for most workloads. It is built for long-running agentic coding and knowledge work: multihour autonomous coding agents, large-scale refactoring, complex systems engineering, vision-heavy workflows and computer use. Its API ID is claude-opus-5-5.

Opus 5.5 is also cheaper than the Opus models before it: $4 input and $20 output per million tokens, versus $5 / $25 for Opus 5 and Opus 4.8. Cache reads cost $0.20 per million tokens, or 5% of the input price. Adaptive thinking is always on and can’t be turned off; the effort parameter controls depth, with medium as the API default. A separately priced fast mode is available on the Claude API; Anthropic’s pricing page describes it as up to 2.5x faster at 2x standard pricing.

Opus 4.8 is now a legacy model. Its fast mode was deprecated on October 30, 2026 and will be removed on November 30, 2026. Background on that release is in our Claude Opus 4.8 status explainer.

Claude Fable 5.2: for the hardest reasoning work

Claude Fable 5.2 is the current Fable model, released on November 9, 2026 and available to all customers on the Claude API and partner platforms as claude-fable-5-2. It costs $8 input and $40 output per million tokens, 20% less than Fable 5.1’s $10 / $50, and its cache reads cost $0.20 per million tokens, the same as Opus 5.5. Anthropic says it produces output up to 30% faster than Fable 5.1 and positions it for demanding reasoning and long-horizon agentic work, or for cases where Opus 5.5 at higher effort still falls short in your evals. Its retirement date is not sooner than November 9, 2027. Our Claude Fable 5.2 guide covers the benchmarks, plan access and migration in detail.

Adaptive thinking is always on. The default thinking.display is omitted; a summarized display is available, and the raw chain of thought is never returned. The API default effort is high. Fable 5.2 carries 30-day data retention and is not available under zero data retention unless Anthropic expressly authorizes it, which matters for regulated or highly sensitive workloads. Code written for Fable 5.1 needs checking before migration: the history-editing check on thinking blocks now applies to every account, and computer_20251124 returns an error everywhere except Amazon Bedrock.

Availability distinction:

Fable 5.2 is the generally available model. Claude Mythos 5.2 is the same underlying model at the same price, without Fable’s cyber and biology safeguards, but only for organizations verified through Anthropic’s verification programs, such as the Cyber Verification Program. On Claude subscriptions, Fable is included on Max and Ultra up to 50% of the weekly limit and is billed through usage credits on Pro, apart from a launch week that ends November 16, 2026; our Fable pricing and limits guide has the details.

Claude Sonnet 5.5: the balanced production model

Claude Sonnet 5.5 is the current Sonnet model, released on September 28, 2026 as claude-sonnet-5-5. Anthropic positions it as the best combination of speed and intelligence for everyday coding, agent and enterprise workloads. It costs $2 input and $10 output per million tokens, the same list price as Sonnet 5, with cache reads at $0.10.

Adaptive thinking is on by default and the API default effort is high. Thinking can’t be fully disabled: the lowest setting is between_tools, which turns off up-front thinking at high effort or below. Migrating from Sonnet 5 involves several breaking changes, which our Sonnet 5.5 pricing and migration guide covers with before-and-after code. For the previous generation, see our analysis of Claude Sonnet 5 for coding agents.

Claude Haiku 5.5: fastest and least expensive

Claude Haiku 5.5 is the current Haiku model, released on October 7, 2026 as claude-haiku-5-5. It is built for high-volume, latency-sensitive work such as classification, routing, extraction and sub-agent tasks. Compared with Haiku 4.5, it jumps from a 200K to a 1M-token context window and from 64K to 128K maximum output.

Haiku 5.5 supports adaptive thinking, on by default, steered by the effort parameter with medium as the default. It uses the newer tokenizer introduced with Claude Opus 4.7, so the same text counts as roughly 30% more tokens than on Haiku 4.5. Include that when you compare its tiered price with Haiku 4.5’s flat $1 / $5. Omit temperature, top_p and top_k from requests.

Legacy and deprecated Claude models

Anthropic still serves several earlier models as legacy: Fable 5.1, Fable 5, Opus 5, Opus 4.8, Opus 4.7, Opus 4.6, Opus 4.5, Sonnet 5, Sonnet 4.6 and Haiku 4.5. They keep working, but new projects should start on the current lineup.

Legacy modelInput / output per 1MCache readNote
Fable 5.1$10 / $50$0.25Replaced by Fable 5.2 at $8 / $40; still active, retirement not sooner than September 1, 2027
Fable 5$10 / $50$1Replaced by Fable 5.1 at the same base price
Opus 5$5 / $25$0.50Replaced by Opus 5.5
Opus 4.8$5 / $25$0.50Fast mode removed November 30, 2026
Sonnet 5$2 / $10$0.20The introductory price became permanent; the planned September 1, 2026 rise to $3 / $15 did not happen
Haiku 4.5$1 / $5$0.10200K context, 64K output

Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) is deprecated and retires on November 30, 2026, with claude-sonnet-5-5 as Anthropic’s recommended replacement. Opus 4.1 retired on August 5, 2026, and Opus 4 and Sonnet 4 retired on June 15, 2026. Anthropic says it gives at least 60 days’ notice before retiring a publicly released model.

Claude pricing comparison

These are standard Claude API token prices published by Anthropic. Cloud-provider pricing, US-only inference (1.1x), batch processing, prompt caching and tool-specific charges can change the effective total.

ModelInput / 1MOutput / 1MCache read / 1MApprox. cost: 10M input + 2M output
Fable 5.2$8$40$0.20$160
Opus 5.5$4$20$0.20$80
Sonnet 5.5$2$10$0.10$40
Haiku 5.5 (prompts up to 100K tokens)$0.10$0.50$0.01$2
Haiku 5.5 (prompts over 100K tokens)$0.50$2.50$0.05$10

The example multiplies list prices by 10 million input tokens and 2 million output tokens spread across many requests. It excludes caching, batch discounts, thinking-token consumption and provider-specific charges. Haiku 5.5’s rate depends on the size of each prompt, not on monthly volume. Run the calculation again with measured token counts from your own prompts because output length and adaptive thinking can dominate cost.

Context, thinking and API migration details

ModelThinkingDefault effort (API)Context / max outputReliable knowledge cutoff
Fable 5.2Adaptive, always onhigh1M / 128KJune 2026
Opus 5.5Adaptive, always onmedium1M / 128KJune 2026
Sonnet 5.5Adaptive, on by default; lowest setting between_toolshigh1M / 128KJune 2026
Haiku 5.5Adaptive, on by defaultmedium1M / 128KJune 2026
  • Model IDs are pinned. Anthropic says every model ID is a pinned snapshot, including the dateless IDs used since the 4.6 generation.
  • Output limits include thinking. Thinking tokens count toward max_tokens, so leave enough room for both reasoning and the final answer. On the Batch API, Opus 5.5, Sonnet 5.5 and Haiku 5.5 can return up to 300K output tokens with a beta header.
  • Manual thinking budgets are gone. Extended thinking with budget_tokens is not accepted on the current models; use adaptive thinking and effort.
  • Sampling parameters are removed. temperature, top_p and top_k return a 400 error in API version 2026-10-21; earlier API versions reject non-default values on Claude 4.7 and later models.
  • Newer tokenization changes comparisons. Claude 4.7 and later models produce roughly 30% more tokens for the same text than Sonnet 4.6, Haiku 4.5 and older models.
  • Benchmark with real data. Capability labels are a selection guide, not a substitute for evaluations of accuracy, tool use, latency, refusals and total cost.

Practical recommendation

Use Haiku 5.5 for cheap, fast, high-volume steps; Sonnet 5.5 for most everyday execution where cost matters; Opus 5.5 as the default for long-running agentic coding and knowledge work; and Fable 5.2 when your evals show Opus 5.5 at higher effort still falls short and the doubled token price and 30-day retention constraint are acceptable.

A robust production system can route tasks between tiers: Haiku for triage, Sonnet or Opus for the majority of execution, and Fable for escalations. Track quality and cost per completed task, not just cost per token.

Looking for the older leaked names rather than today’s lineup? The Claude Sonnet 4.8 and Capybara historical archive records what shipped, changed names or remained unconfirmed without treating the leak as a current model guide.

Frequently asked questions

What are the current Claude models in 2026?

As of November 12, 2026, Anthropic’s current lineup is Claude Fable 5.2, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5. Fable 5.1, Fable 5, Opus 5, Opus 4.8, Sonnet 5 and Haiku 4.5 remain available as legacy models.

What is the most capable Claude model?

Claude Fable 5.2, released on November 9, 2026, is Anthropic’s most capable generally available model, built for demanding reasoning and long-horizon agentic work. Claude Mythos 5.2 is the same underlying model without Fable’s cyber and biology safeguards and is limited to verified organizations. Anthropic recommends starting with Claude Opus 5.5 for most workloads.

Is Claude Fable 5.2 generally available?

Yes. Claude Fable 5.2 is available to all customers on the Claude API and partner platforms, and on all paid Claude plans. Claude Mythos 5.2 is offered only to organizations verified through Anthropic’s verification programs, such as the Cyber Verification Program. Fable 5.1 remains available and will not retire before September 1, 2027.

How do Opus 5.5, Fable 5.2 and Sonnet 5.5 differ?

Opus 5.5 is the default for long-running agentic coding and knowledge work, Fable 5.2 targets the most demanding reasoning and long-horizon agents at twice Opus 5.5’s token price, and Sonnet 5.5 offers the best balance of speed and intelligence at half Opus 5.5’s token price.

How much do current Claude models cost?

Per million API tokens, Fable 5.2 is $8 input and $40 output, Opus 5.5 is $4 and $20, Sonnet 5.5 is $2 and $10, and Haiku 5.5 is $0.10 and $0.50 for prompts up to 100,000 tokens or $0.50 and $2.50 above that.

Which Claude models have a 1M-token context window?

All four current models, Fable 5.2, Opus 5.5, Sonnet 5.5 and Haiku 5.5, have a 1M-token context window and up to 128K output tokens. The legacy Claude Haiku 4.5 has 200K context and up to 64K output.

Which Claude model is best for coding?

Anthropic recommends Opus 5.5 as the starting point for long-running agentic coding. Sonnet 5.5 is a cheaper option for everyday coding, and Fable 5.2 is for cases where Opus 5.5 at higher effort still falls short. Confirm the choice with your own coding evaluations.

Update log

  • November 12, 2026: Replaced Fable 5.1 with Claude Fable 5.2 (released November 9, 2026, $8 / $40, cache reads $0.20, Moderate latency, retirement not sooner than November 9, 2027); moved Fable 5.1 to the legacy table; replaced Mythos 5.1 with Mythos 5.2; recalculated the 10M input + 2M output example.
  • November 7, 2026: Replaced the lineup with Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5; added Haiku 5.5’s tiered pricing, current cache-read prices, thinking and effort defaults, a legacy and deprecated models section, and the removal of sampling parameters in API version 2026-10-21. Removed the expired Sonnet 5 introductory-price warning (its $2 / $10 price became permanent).
  • July 18, 2026: Rebuilt the comparison against Anthropic’s models overview; added Fable 5 general availability, Sonnet 5, pricing and thinking-mode behavior.
  • Verification rule: Model status, IDs, prices and limits are taken from official Anthropic documentation linked below. Recheck time-sensitive prices before budgeting.
Official sources
  1. Anthropic — Models overview. Current model lineup, IDs, pricing, latency, context, thinking, effort defaults and legacy models.
  2. Anthropic — Claude API pricing. Base rates, cache reads and writes, Haiku 5.5 tiers and the permanent Sonnet 5 price.
  3. Anthropic — Choosing the right model. Model-selection guidance and evaluation advice.
  4. Anthropic — Claude Opus 5.5 overview. Release date, positioning, pricing, thinking and fast mode.
  5. Anthropic — Introducing Claude Fable 5.2 and Claude Mythos 5.2. Release date, output speed, safeguards and Mythos 5.2 access.
  6. Anthropic — Claude Fable 5.2 overview. Positioning, pricing, context and availability.
  7. Anthropic — What’s new in Claude Fable 5.2. Breaking changes, pricing, cache reads and fast mode.
  8. Anthropic — Claude Fable 5.1 overview. Release date, pricing, availability and Mythos 5.1 distinction.
  9. Anthropic — What’s new in Claude Fable 5.1. Breaking changes, thinking display, data retention and cache pricing.
  10. Anthropic — Claude Sonnet 5.5 overview. Release date, pricing, thinking settings and limits.
  11. Anthropic — Claude Haiku 5.5 overview. Release date, tiered pricing, context, tokenizer and thinking.
  12. Anthropic — Model deprecations. Lifecycle status, retirement dates (including Fable 5.2 and Fable 5.1) and the sampling-parameter removal.
  13. Anthropic — Claude plans and pricing. Subscription plans, Fable inclusion and fast-mode pricing.
  14. Anthropic — Claude Code fast mode. Opus 4.8 fast-mode deprecation and removal dates.
  15. Anthropic — Claude Fable models on your plan. Fable inclusion on Max, Ultra and Team premium seats; usage credits on Pro.
  16. Anthropic — Effort. Reasoning-depth controls for adaptive-thinking models.
  17. Anthropic — Context windows. Context-window behavior and limits.
  18. Anthropic — Model IDs and versioning. Pinned snapshots, aliases and naming.
RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

- Advertisment -

Most Popular

Recent Comments