As of July 2026 there are six Claude models in active service: Claude Fable 5 (the new flagship), Opus 4.8 and 4.7, Sonnet 5 and 4.6, and Haiku 4.5. This guide explains what each one is for, what it costs, and how the tiers relate — so you stop guessing at the model picker and start choosing deliberately.
TL;DR:
• Three tiers: Haiku (fast, cheap), Sonnet (balanced), Opus (most capable) — plus Fable 5 sitting above Opus as the new flagship line.
• Claude Fable 5 is Anthropic’s most intelligent generally available model, with thinking always on.
• Opus 4.8 is the default recommendation for demanding work; Sonnet 5 is the best speed/intelligence balance.
• Current Opus, Sonnet, and Fable models have a 1M-token context window; Haiku 4.5 has 200K.
• API prices per million tokens: Fable 5 $10/$50, Opus 4.8 $5/$25, Sonnet 5 $3/$15, Haiku 4.5 $1/$5.
All current Claude models at a glance
Every model below is generally available through the claude.ai apps (on paid plans) and the API. Prices are API rates per million tokens, input/output.
| Model | API ID | Price in/out (MTok) | Context | Max output | Best for |
|---|---|---|---|---|---|
| Claude Fable 5 | claude-fable-5 | $10 / $50 | 1M | 128K | Hardest reasoning and agentic work |
| Claude Opus 4.8 | claude-opus-4-8 | $5 / $25 | 1M | 128K | Demanding coding, analysis, long agents |
| Claude Opus 4.7 | claude-opus-4-7 | $5 / $25 | 1M | 128K | Previous-gen Opus, still active |
| Claude Sonnet 5 | claude-sonnet-5 | $3 / $15 (intro $2 / $10 to Aug 31, 2026) | 1M | 128K | Everyday work, near-Opus coding |
| Claude Sonnet 4.6 | claude-sonnet-4-6 | $3 / $15 | 1M | 128K | Previous-gen Sonnet, still active |
| Claude Haiku 4.5 | claude-haiku-4-5 | $1 / $5 | 200K | 64K | High volume, low latency, simple tasks |
How the Claude model tiers work
Anthropic names its Claude models after poetry forms, smallest to largest: Haiku is the fast, cheap tier; Sonnet is the balanced middle; Opus is the heavyweight. The number after the name is the generation (4.5, 4.8, 5), and generations overlap — Opus 4.7 stayed in service after 4.8 shipped, which is why the lineup always has more entries than you would expect. In 2026 Anthropic added a fourth name on top: Fable, the first of the Claude 5 family, positioned above Opus.
One spec quietly became the great equalizer: context. The 200K-token window was the long-time Claude standard, but every current Opus, Sonnet, and Fable model now takes 1M tokens — several long novels’ worth of input — while Haiku 4.5 remains at 200K. Our context window explainer covers what those numbers mean in practice.
Claude Fable 5 — the flagship
Claude Fable 5 is Anthropic’s most intelligent generally available model and the first of the Claude 5 family. Two things make it structurally different from the rest of the lineup. First, thinking is always on — you don’t toggle extended reasoning, the model decides how hard to think. Second, it sits above Opus in capability and in price: at $10/$50 per million tokens it costs twice Opus 4.8 on input and output alike.
In my testing, Fable 5 earns that premium on genuinely hard problems — multi-step agentic tasks, subtle architectural refactors, research synthesis where the reasoning chain matters. For routine work the difference over Opus 4.8 is rarely worth 2× the cost. One footnote for completeness: Claude Mythos 5 is the same underlying model as Fable 5, offered to approved organizations only, without Fable’s additional dual-use safety measures. Ordinary users and developers will never touch it.
Claude Opus 4.8 and 4.7 — the workhorses for hard problems
Opus 4.8 is the most capable Opus-tier model and Anthropic’s default recommendation for demanding work: large codebases, long-horizon agents, serious analysis. It also supports fast mode, which runs the model at higher output speed — in Claude Code that’s the /fast command. Opus 4.7, the previous generation, remains active at the same $5/$25 price; teams with validated pipelines on it have little reason to rush a migration, but new projects should start on 4.8.
Claude Sonnet 5 and 4.6 — the default for everyday work
Sonnet 5 is the model I leave selected most of the day. It offers the best speed/intelligence balance in the lineup and gets near-Opus quality on coding at 60% of the price — and through August 31, 2026 there’s an intro API rate of $2/$10 that makes it an even easier default. Sonnet 4.6 stays available at the standard $3/$15 for workloads already tuned to it. If you’re weighing the middle tier against the top one seriously, our deep technical comparison of Opus vs Sonnet runs the two head to head, capability by capability.
Claude Haiku 4.5 — fast and cheap
Haiku 4.5 is the speed tier: $1/$5 per million tokens, a 200K context window, and 64K max output. It’s the right engine for classification, extraction, routing, summarization at volume, and latency-sensitive user-facing features. It will disappoint you on multi-step reasoning — that’s not its job. A common production pattern is Haiku triaging requests and escalating the hard ones to Sonnet or Opus.
Don’t dismiss it as a toy, though. In one of my pipelines Haiku 4.5 handles tens of thousands of short extraction calls a day at a cost Sonnet would multiply by three — and for that shape of work its answers are indistinguishable from the bigger models’.
Which Claude model should you choose?
- Default for most people: Sonnet 5 — fast, capable, and the cheapest of the current-gen mainline models.
- Demanding coding and analysis: Opus 4.8, with fast mode when you want more speed.
- The absolute hardest problems: Fable 5, when the outcome justifies 2× Opus pricing.
- High-volume or real-time API work: Haiku 4.5.
In the claude.ai apps, all of these come with a paid plan and the picker sits at the bottom of the chat box; the specific features each plan wraps around the models are covered in our Claude AI features guide. On the API, model choice is a per-request parameter — full rate tables live in our Claude API pricing breakdown and on Anthropic’s official pricing page.
Common mistakes when picking between Claude models
- Defaulting to the biggest model for everything. Fable 5 on routine summarization is money spent on reasoning you never use.
- Judging a tier by an old generation. “Sonnet isn’t good enough for coding” was a 2024 opinion; Sonnet 5 is near-Opus on code.
- Forgetting Haiku’s different limits. It’s the one current model still at 200K context and 64K output — plan long-document pipelines accordingly.
- Ignoring the Sonnet 5 intro pricing. Until August 31, 2026, $2/$10 makes it cheaper than some teams’ Haiku-era budgets assumed.
Next steps
Pick the tier that matches your hardest regular task, not your hardest imaginable one — you can always escalate a single conversation to a bigger model. For a structured way to decide between the two tiers most people actually agonize over, read our guide section pillar or jump straight into the Opus vs Sonnet comparison linked above.
FAQ
What Claude models are available in 2026?
Six models are current: Claude Fable 5, Opus 4.8, Opus 4.7, Sonnet 5, Sonnet 4.6, and Haiku 4.5. Fable 5 is the flagship, Opus is the high-capability tier, Sonnet is the balanced tier, and Haiku is the fast, low-cost tier.
What is the difference between Claude Opus, Sonnet, and Haiku?
They are capability tiers named after poetry forms. Haiku is the fastest and cheapest, Sonnet balances speed and intelligence for everyday work, and Opus is the most capable of the three for demanding tasks. Fable 5 now sits above Opus as the flagship.
What is Claude Fable 5?
Claude Fable 5 is Anthropic’s newest flagship and the first model of the Claude 5 family. It is the company’s most intelligent generally available model, has thinking always on, a 1M-token context window, and costs $10 per million input tokens and $50 per million output tokens on the API.
Which Claude model is best for coding?
Sonnet 5 is the best default for most coding work, delivering near-Opus quality at $3/$15 per million tokens. For large codebases, tricky debugging, or long agentic sessions, Opus 4.8 is the recommended step up, and Fable 5 handles the hardest problems.
Do all Claude models have a 1M context window?
No. Current Opus, Sonnet, and Fable models support 1M tokens of context, but Haiku 4.5 remains at the older 200K standard and also has a smaller 64K maximum output.
What is Claude Mythos 5?
Mythos 5 is the same underlying model as Fable 5 but offered without Fable’s additional dual-use safety measures, and only to approved organizations. It is not available to regular consumers or standard API customers.