Best models

Best AI models for structured output and data extraction

Every model in the live reference catalogue with structured output, ranked by the intelligence index (Artificial Analysis, 0-100) with vendor list prices per million tokens. 12 qualify today.

Strongest

Claude Opus 5.5

intelligence index 58 · $8.00 blended

Best value

Muse Spark 1.3

48 points for $2.00

Cheapest

Muse Spark 1.3

$2.00 blended · index 48

Ranked by the intelligence index

Strongest first; price breaks ties. Value is index points per blended dollar. Click a model for its full specification, or compare two side by side.

#	Model	Intelligence index	Input / 1M	Output / 1M	Context	Value	Compare
1	Claude Opus 5.5Anthropic	58	$4.00	$20.0	1M	7
2	Claude Sonnet 5.5Anthropic	56	$2.00	$10.0	1M	14	vs Claude Opus 5.5
3	Claude Fable 5.1Anthropic	53	$10.0	$50.0	1M	3	vs Claude Opus 5.5
4	GPT-6 AstraOpenAI	53	$10.0	$50.0	1.1M	3	vs Claude Opus 5.5
5	Gemini 4 ArgonGoogle	53	$2.00	$10.0	1.0M	13	vs Claude Opus 5.5
6	GPT-6.1 SolOpenAI	52	$2.00	$10.0	1.1M	13	vs Claude Opus 5.5
7	Claude Opus 5Anthropic	51	$5.00	$25.0	1M	5	vs Claude Opus 5.5
8	Claude Fable 5Anthropic	50	$10.0	$50.0	1M	2	vs Claude Opus 5.5
9	Muse Spark 1.3Meta	48	$1.25	$4.25	1.0M	24	vs Claude Opus 5.5
10	GPT-6 SolOpenAI	48	$2.00	$10.0	1.1M	12	vs Claude Opus 5.5
11	GPT-5.6 SolOpenAI	47	$4.00	$20.0	1.1M	6	vs Claude Opus 5.5
12	Grok 4.7xAI	46	$2.00	$6.00	500K	15	vs Claude Opus 5.5

What to weigh for structured output and extraction

When a program parses the answer, the model's job is to produce exactly the shape asked for - a JSON object matching a schema, a label from a fixed set, a table. Only models offering structured (schema-constrained) output qualify here; a model that is merely asked nicely for JSON returns prose often enough to break a pipeline.

Extraction and classification are high-volume and short, so price matters as much as the index - the value ranking below is usually the right one. The index still earns its place: a model that constrains the shape but misreads the document fills the schema with wrong values.

The router adds two things. It only considers models that honour the schema when the request carries one, and it learns from your outcomes - a parse failure or a rejected extraction reported as feedback moves the next similar request to a model that holds up. Extraction pipelines also benefit from the output guard: a response that does not match the schema is caught before it reaches your code, retried on the next admissible model, and counted against the first model's quality estimate so the ranking the router uses reflects what actually parsed.

The rule, in OpenSmartRoute

Declare the models as targets and let the rule send the hard requests to the strongest one and the rest to the best value; the trace names the rule that fired.

yaml

# rules.yaml - structured-output: the strongest qualifying model for the hard requests, the best value for the rest
rules:
  - name: structured-output-hard
    when: { actions: [extract, classify] }
    prefer: [claude-opus-5.5]
    weight: 0.8
  - name: structured-output-default
    when: { max_complexity: 0.45 }
    prefer: [muse-spark-1.3]
    weight: 0.6

Try a request Price a workload

Frequently asked

Which models guarantee valid JSON?: The ones in the ranking above, which all publish schema-constrained output; the catalogue flags it per model and the router only matches a schema request to them.
Is the cheapest structured-output model good enough?: For classification and simple extraction, often. The value column ranks index per dollar; try the top value model first and let outcomes decide.
What if the model returns the wrong values?: Report the outcome; the router treats it like any bad answer and shifts similar requests to a stronger qualifying model.

Best models for other tasks

For the business

What it means for which models you pay for.

Catalogue of every vendor
The right model, not the famous one
Price, context, abilities and benchmark indexes side by side, with the model that fits each task named.
The rankings
89% less
Lower cost per request
On the 150-prompt benchmark the router spent 89% less than sending every prompt to the reference model, and the judged quality went up, not down.
Read the benchmark
Per million tokens, every model
Prices you can compare
The catalogue prints input and output prices and what three typical monthly workloads cost on each.
The model catalogue

Which LLM is the cheapest for our workload?: The best-for pages rank the live catalogue per task by the index that measures it, and the router picks the cheapest model that meets your quality rule for each request.
How much can we save on our LLM bill?: On the public benchmark the router saved 89% against sending everything to the reference model. Import last month's provider bill on the bill page and the estimate is computed on your own traffic before you change a line of code.
How often do model prices change?: The catalogue is refreshed from the vendors' price lists; each model page shows the date of its data.

Let the router pick the structured output and extraction model per request.

The ranking above is where to start; the router finishes the job - the strongest qualifying model for the hard requests, the best value for the rest, and every decision explained.

Create a free account Price a workload

Free plan, no card. Fifteen thousand decisions a month with the full trace.

Model

Intelligence index

Input / 1M

Output / 1M

Context

Value

Compare

Claude Opus 5.5Anthropic

$4.00

$20.0

Claude Sonnet 5.5Anthropic

$2.00

$10.0

vs Claude Opus 5.5

Claude Fable 5.1Anthropic

$10.0

$50.0

vs Claude Opus 5.5

GPT-6 AstraOpenAI

$10.0

$50.0

1.1M

vs Claude Opus 5.5

Gemini 4 ArgonGoogle

$2.00

$10.0

1.0M

vs Claude Opus 5.5

GPT-6.1 SolOpenAI

$2.00

$10.0

1.1M

vs Claude Opus 5.5

Claude Opus 5Anthropic

$5.00

$25.0

vs Claude Opus 5.5

Claude Fable 5Anthropic

$10.0

$50.0

vs Claude Opus 5.5

Muse Spark 1.3Meta

$1.25

$4.25

1.0M

vs Claude Opus 5.5

GPT-6 SolOpenAI

$2.00

$10.0

1.1M

vs Claude Opus 5.5

GPT-5.6 SolOpenAI

$4.00

$20.0

1.1M

vs Claude Opus 5.5

Grok 4.7xAI

$2.00

$6.00

500K

vs Claude Opus 5.5

What to weigh for structured output and extraction

The rule, in OpenSmartRoute

Declare the models as targets and let the rule send the hard requests to the strongest one and the rest to the best value; the trace names the rule that fired.

yaml

# rules.yaml - structured-output: the strongest qualifying model for the hard requests, the best value for the rest
rules:
  - name: structured-output-hard
    when: { actions: [extract, classify] }
    prefer: [claude-opus-5.5]
    weight: 0.8
  - name: structured-output-default
    when: { max_complexity: 0.45 }
    prefer: [muse-spark-1.3]
    weight: 0.6

Frequently asked

Which models guarantee valid JSON?

The ones in the ranking above, which all publish schema-constrained output; the catalogue flags it per model and the router only matches a schema request to them.

Is the cheapest structured-output model good enough?

For classification and simple extraction, often. The value column ranks index per dollar; try the top value model first and let outcomes decide.

What if the model returns the wrong values?

Report the outcome; the router treats it like any bad answer and shifts similar requests to a stronger qualifying model.

Best AI models for structured output and data extraction

Ranked by the intelligence index

What to weigh for structured output and extraction

The rule, in OpenSmartRoute

Frequently asked

Related

Best models for other tasks

What it means for which models you pay for.

The right model, not the famous one

Lower cost per request

Prices you can compare

Let the router pick the structured output and extraction model per request.

Best AI models for structured output and data extraction

Ranked by the intelligence index

What to weigh for structured output and extraction

The rule, in OpenSmartRoute

Frequently asked

Related

Best models for other tasks

What it means for which models you pay for.

The right model, not the famous one

Lower cost per request

Prices you can compare

Let the router pick the structured output and extraction model per request.