Anthropic's Claude lineup in mid-2026 runs four tiers from fastest to most capable: Haiku for speed and volume, Sonnet for everyday work, Opus for the hardest reasoning, and Fable as the new top tier above all three. The simple rule is to start with Sonnet, which now means the freshly launched Sonnet 5, and only step up to Opus when a task genuinely needs more accuracy, down to Haiku when you need cheap speed at scale, or up to Fable for frontier-level work. One naming note that trips people up: there is no model called 'Claude 5.' Fable and Mythos are a new tier, not a version number.
Picking the wrong tier quietly costs you, either in money on an overkill model or in quality on an underpowered one. This guide explains what each Claude model is actually for, where Sonnet 5 fits after its June 30 launch, the real prices, and a clean way to choose without memorizing benchmarks.
Quick Answer: The Claude Lineup at a Glance
Four tiers, four jobs. Prices are per million tokens for input and output on the Claude Platform.
| Model | Best for | Price (in / out) |
|---|---|---|
| Haiku 4.5 | Speed and high volume: classification, routing, simple tasks | $1 / $5 |
| Sonnet 5 | The default daily driver: writing, analysis, coding, agents | $3 / $15 |
| Opus 4.8 | Hardest reasoning, long autonomous coding, browser agents | $5 / $25 |
| Fable 5 | Frontier-level work, the most capable public tier | $10 / $50 |
Haiku 4.5: Fast and Cheap, On Purpose
Haiku is Anthropic's fastest and most affordable model, priced at roughly $1 input and $5 output per million tokens. It is built for jobs where speed and cost matter more than deep reasoning: classifying or routing messages, extracting fields from text, powering high-volume chat, and pre-processing before a bigger model takes over. With a 200K-token context window it handles short-to-medium tasks comfortably; for very long documents or hard reasoning, step up to Sonnet or Opus. If you are running something thousands of times, Haiku is usually the cost-smart choice.
Sonnet 5: The New Default for Most People
Sonnet is the balanced middle, and as of June 30, 2026 that means Sonnet 5, the most agentic Sonnet Anthropic has shipped. It performs close to the pricier Opus 4.8 on many tasks while staying at Sonnet pricing of $3 input and $15 output per million tokens, and it is the default model for Free and Pro users. For everyday writing, analysis, coding, and tool-using agents, it is the right starting point for almost everyone. Anthropic also reports it hallucinates and flatters less than the previous Sonnet 4.6, which matters for trusting its answers.
Opus 4.8: The Heavy Lifter
Opus is the flagship for difficulty. At about $5 input and $25 output per million tokens, it is the model to reach for when a task needs the highest accuracy: complex multi-step reasoning, long autonomous coding sessions, demanding browser and computer-use agents, and work where being right is worth the higher cost. Anthropic also points security researchers to Opus 4.8 for cybersecurity work that needs lighter guardrails. The honest framing is not 'Opus is better, always,' it is 'Opus earns its price on the hardest jobs and is overkill for routine ones.'
Fable 5: The New Top Tier (and the Naming Change)
In June 2026 Anthropic shipped Fable 5, its most capable widely released model and the first of a new Mythos-class tier that sits above Opus, Sonnet, and Haiku. At roughly $10 input and $50 output per million tokens it targets frontier-level reasoning and long-horizon agentic work, with safety classifiers that route high-risk requests in areas like cybersecurity and biology to Opus 4.8 instead. Two things to know: there is a research-only sibling, Mythos 5, available by invitation through Anthropic's Project Glasswing, and access to the Fable and Mythos tier was temporarily suspended in June 2026 under a US export-control directive, so availability can change, and Anthropic's announcements are the place to check.
The single most common Claude confusion: there is no 'Claude 5.' Anthropic changed its naming, so Fable and Mythos are a new capability tier above Opus, Sonnet, and Haiku, not the next version number. Sonnet 5 is simply the fifth-generation Sonnet, the mid tier.
How to Pick in One Step
- Default to Sonnet 5. For most writing, analysis, coding, and agent work it is the right balance of quality and cost.
- Drop to Haiku 4.5 when you are running something at high volume and the task is simple, like sorting or extracting.
- Step up to Opus 4.8 only when a task is genuinely hard and accuracy is worth the higher price, or for security work needing lighter guardrails.
- Consider Fable 5 for frontier-level work, checking current availability first since the tier's access can be restricted.
- Do not pay for Opus or Fable on tasks Sonnet handles well; the overkill is pure wasted cost.
Comparing Claude Tiers Without Three Accounts
The best way to feel the difference between tiers is to run the same prompt across a few and see where the cheaper model already wins. That is hard when each sits behind its own plan, which is where a multi-model tool helps: LumiChats puts Claude alongside 40-plus models in one place, so you can test whether a task really needs Opus or whether Sonnet 5 nails it, and route your work to the tier that earns its cost rather than defaulting to the most expensive one.
01What is the difference between Claude Haiku, Sonnet, and Opus?
Haiku is the fastest and cheapest, built for high-volume simple tasks. Sonnet is the balanced default for everyday writing, analysis, and coding, now meaning Sonnet 5. Opus is the flagship for the hardest reasoning and most demanding agentic work, at the highest price of the three classic tiers.
02Which Claude model should I use?
Start with Sonnet 5, the default for most people, since it balances quality and cost for everyday work. Use Haiku 4.5 for cheap, high-volume simple tasks, and step up to Opus 4.8 only when a task genuinely needs the highest accuracy. Fable 5 is the frontier tier for the most demanding work.
03Is there a Claude 5?
No. Anthropic changed its naming, so there is no model called 'Claude 5.' Fable and Mythos form a new capability tier above Opus, Sonnet, and Haiku. Sonnet 5 is the fifth generation of the mid-tier Sonnet model, not an overall 'Claude 5.'
04What is Claude Fable 5?
Fable 5 is Anthropic's most capable widely released model and the first public model in its Mythos-class tier, above Opus, Sonnet, and Haiku. It targets frontier reasoning and long-horizon agentic work, with safety routing for high-risk requests. Access to the Fable and Mythos tier can be restricted, so check Anthropic's announcements for current availability.
05Which Claude model is free?
Free Claude users get the current default model, which as of June 30, 2026 is Sonnet 5. The free tier has usage limits, while paid plans like Pro, Max, Team, and Enterprise raise those limits and add access to higher tiers such as Opus 4.8.
The lineup looks complex but the choice is simple: Sonnet 5 by default, Haiku for cheap volume, Opus for the hardest work, Fable for the frontier, and no 'Claude 5' to worry about. Match the tier to the task instead of always reaching for the most powerful one, and you get the best of Claude without overpaying for it.
