Business Strategy & ROIRole-Specific

Which Claude model should your team use?

In brief

Claude is a four-model lineup now — Haiku 4.5 at $1/$5, Sonnet 5 at $2/$10, Opus 5 at $5/$25, Fable 5 at $10/$50 — and the gaps are narrower than they used to be. How to pick, why Sonnet 5 should be your default, and why tuning the effort parameter usually beats downgrading the model.

5 min read·Claude Plans

Contents

♡Sign in to save

Claude comes in four models: Fable 5 (highest capability), Opus 5 (complex agentic and enterprise work), Sonnet 5 (the balance of speed and intelligence), and Haiku 4.5 (fastest, cheapest). Most teams default to whatever sounds most powerful. That wastes money and often produces slower results without meaningful quality improvement.

Here is how to actually choose.

The current lineup

Model	Price (input / output per Mtok)	Context	Best for
Claude Haiku 4.5	$1 / $5	200K	Classification, extraction, high-volume routine work
Claude Sonnet 5	$2 / $10	1M	The daily driver — most work belongs here
Claude Opus 5	$5 / $25	1M	Complex agentic coding, enterprise work
Claude Fable 5	$10 / $50	1M	Long-running agents, the hardest problems

The gaps are narrower than they used to be. Opus 5 is 2.5x Sonnet 5, not the 5x that separated the equivalent tiers a generation ago, and Sonnet 5 costs a third less than Sonnet did before its pricing became permanent in August 2026. Check the pricing page before budgeting — this lineup moves.

What each model is good at

Haiku 4.5 — the workhorse. Fast, cheapest, 200K context. Use for:

Simple Q&A, classification, extraction
Reformatting text, adjusting tone
Sorting, categorising, and tagging
Subagents doing one scoped mechanical job
High-volume tasks where cost scales linearly

Sonnet 5 — the daily driver, and where most teams should spend most of their time. Use for:

Content drafting, summarising, analysis requiring judgment
Code generation and review
Multi-step tasks needing coherent reasoning
Agent work that doesn't need the top tier

Opus 5 — the heavy lifter. Anthropic positions it for complex agentic coding and enterprise work. Use when:

The problem is genuinely hard: ambiguous inputs, nuanced judgment
You're running long agentic loops with many tool calls
Sonnet's output isn't good enough for a specific use case

Fable 5 — next-generation intelligence for long-running agents, with thinking always on and no option to disable it. At $10/$50 it is double Opus 5, so use it when you have measured that Opus 5 leaves capability on the table, not as a default.

The mistake teams make

Defaulting to the top of the lineup for everything. It is like taking a taxi for every trip when most of them are a five-minute walk.

The right approach: start with Sonnet 5. If output quality isn't good enough for a specific use case, upgrade that use case. Keep everything else on Sonnet. Route genuinely simple tasks to Haiku 4.5.

On Claude.ai you select the model per conversation. On the API you specify it per request, which means you can route different task types to different models programmatically.

The lever most teams miss

Model choice is only half of it. The effort parameter — low through max, defaulting to high — controls how many tokens Claude spends on a request, including tool calls. Sonnet 5 at medium effort is often both cheaper and faster than Haiku 4.5 at high, with better quality.

Tune effort per workload before you downgrade a model. It is the cheaper experiment and it frequently makes the downgrade unnecessary. See when to raise and lower effort.

For team admins: model strategy

Set Sonnet 5 as the default. Most people don't need to think about model selection.
Educate on when to upgrade. Deep analysis or mediocre outputs on a specific task — switch that task, not everything.
Watch usage patterns. If limits are getting hit, check whether people are running Opus on work Sonnet handles equally well.
Set effort deliberately for anything automated. Every agent and batch job should have a chosen effort level, not an inherited default.

The simple rule

If in doubt, use Sonnet 5. Upgrade to Opus 5 when the output matters a lot and Sonnet isn't cutting it. Reserve Fable 5 for cases where you have evidence Opus 5 isn't enough. Drop to Haiku 4.5 when the task is simple and you need speed or volume.