Best models

Best AI models for customer support

Every model in the live reference catalogue with tool calls, a blended list price under $3 per million tokens, ranked by the intelligence index (Artificial Analysis, 0-100) with vendor list prices per million tokens. 12 qualify today.

Strongest

Muse Spark 1.3

intelligence index 48 · $2.00 blended

Best value

Ling 3.1 Flash

41 points for $0.11

Cheapest

Ling 3.1 Flash

$0.11 blended · index 41

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	Muse Spark 1.3Meta	48	$1.25	$4.25	1.0M	24
2	Grok 4.7xAI	46	$2.00	$6.00	500K	15	vs Muse Spark 1.3
3	MiMo-V2.6-ProXiaomi	46	$0.43	$0.87	1.1M	85	vs Muse Spark 1.3
4	Qwen3.8 Max (0902)Qwen	45	$2.00	$6.00	1M	15	vs Muse Spark 1.3
5	GLM 5.3Z.ai	45	$0.08	$1.40	1.0M	109	vs Muse Spark 1.3
6	Grok 4.6xAI	44	$2.00	$6.00	500K	15	vs Muse Spark 1.3
7	Step 5 PreviewStepFun	44	$1.00	$2.70	1.0M	31	vs Muse Spark 1.3
8	Claude Haiku 5.5Anthropic	43	$0.10	$0.50	1M	217	vs Muse Spark 1.3
9	GLM 5.3 FlashZ.ai	42	$0.15	$0.50	1.0M	176	vs Muse Spark 1.3
10	Ling 3.1 FlashinclusionAI	41	$0.07	$0.23	262K	365	vs Muse Spark 1.3
11	Gemini 3.8 FlashGoogle	41	$0.82	$4.13	1.0M	25	vs Muse Spark 1.3
12	Qwen3.8 2.4T A95BQwen	40	$2.00	$6.00	1.0M	13	vs Muse Spark 1.3

What to weigh for customer support

A support assistant needs three things: to answer accurately from your knowledge, to look things up - an order, an account, a ticket - through tool calls, and to know when to stop and hand over to a person. Tool calls are required here and the price is capped at $3 blended because support traffic is high-volume and most of it is routine.

The index separates the models that follow a policy from the ones that improvise. A support model that ignores 'do not promise refunds' costs more than its tokens. Within the band, the better-scoring models follow instructions more reliably, which is the quality that matters on a help desk.

The router adds the hand-over. A request that reads as an escalation - anger, a legal word, a request for a human - is pinned to a human queue instead of a model; a request carrying personal data is pinned to the target your data policy allows; and the rest go to the cheapest model that holds up, measured by the outcomes your agents report. Each of those is one rule, and the rules are tried in order, so the escalation and the data boundary always win over the cost preference.

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 - customer-support: the strongest qualifying model for the hard requests, the best value for the rest
rules:
  - name: customer-support-hard
    when: { domains: [customer_support] }
    prefer: [muse-spark-1.3]
    weight: 0.8
  - name: customer-support-default
    when: { max_complexity: 0.45 }
    prefer: [ling-3.1-flash]
    weight: 0.6

Try a request Price a workload

Frequently asked

Which AI model is best for a customer service chatbot?: The ranking above lists the models under $3 blended that support tool calls, strongest first by the intelligence index; the top rows are the usual choice.
How does the assistant hand over to a person?: A rule pins escalation intents to a human target; the router sends the conversation there with its context instead of to a model.
What about customer data in the messages?: The router detects personal data as a signal and a rule pins such requests to the model your policy allows - on-premises or a provider with the right terms.

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 customer support 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

Muse Spark 1.3Meta

$1.25

$4.25

1.0M

Grok 4.7xAI

$2.00

$6.00

500K

vs Muse Spark 1.3

MiMo-V2.6-ProXiaomi

$0.43

$0.87

1.1M

vs Muse Spark 1.3

Qwen3.8 Max (0902)Qwen

$2.00

$6.00

vs Muse Spark 1.3

GLM 5.3Z.ai

$0.08

$1.40

1.0M

109

vs Muse Spark 1.3

Grok 4.6xAI

$2.00

$6.00

500K

vs Muse Spark 1.3

Step 5 PreviewStepFun

$1.00

$2.70

1.0M

vs Muse Spark 1.3

Claude Haiku 5.5Anthropic

$0.10

$0.50

217

vs Muse Spark 1.3

GLM 5.3 FlashZ.ai

$0.15

$0.50

1.0M

176

vs Muse Spark 1.3

Ling 3.1 FlashinclusionAI

$0.07

$0.23

262K

365

vs Muse Spark 1.3

Gemini 3.8 FlashGoogle

$0.82

$4.13

1.0M

vs Muse Spark 1.3

Qwen3.8 2.4T A95BQwen

$2.00

$6.00

1.0M

vs Muse Spark 1.3

What to weigh for customer support

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 - customer-support: the strongest qualifying model for the hard requests, the best value for the rest
rules:
  - name: customer-support-hard
    when: { domains: [customer_support] }
    prefer: [muse-spark-1.3]
    weight: 0.8
  - name: customer-support-default
    when: { max_complexity: 0.45 }
    prefer: [ling-3.1-flash]
    weight: 0.6

Frequently asked

Which AI model is best for a customer service chatbot?

The ranking above lists the models under $3 blended that support tool calls, strongest first by the intelligence index; the top rows are the usual choice.

How does the assistant hand over to a person?

A rule pins escalation intents to a human target; the router sends the conversation there with its context instead of to a model.

What about customer data in the messages?

The router detects personal data as a signal and a rule pins such requests to the model your policy allows - on-premises or a provider with the right terms.

Best AI models for customer support

Ranked by the intelligence index

What to weigh for customer support

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 customer support model per request.

Best AI models for customer support

Ranked by the intelligence index

What to weigh for customer support

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 customer support model per request.