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.

90 days of the enterprise, pro and team plans free, no card needed. See the plans

OpenSmartRoute

Models

Models and routing targets

Everything this deployment can send a request to - 28 targets: models, agents, skills, tools and a human queue, 19 of them answering here. Each row shows the model behind the target, the price per million tokens, the latency measured here and its traffic over the last 7 days. The reference catalogue of 468 hosted models is the second tab.

Routing targetsthis deployment: 28
Answer requests herea provider is attached: 19
Requests routedlast 7 days · this deployment: 591

Vendors behind this deployment

BHS

Recommended per use case

What the router picks today for six common requests, quoted against this catalogue every minute: the pick, what it costs, and the cheapest, best and fastest alternatives. Quote your own prompt on the cost estimator.

Everyday assistant

17 candidates

Mid-tier model

gpt-oss-20bllm

per request: $0.000036
routing quality: Routing quality score: the router's own estimate of how well this target would answer this prompt (0-1), from its quality prior and learned outcomes - not an independent benchmark.

28 of 28

Target	Status	Runs as	Price / 1M	Latency	Tokens 7d	Quality prior
Small writing modelllmwriter-small	answers here	qwen3.5:4bvia ollama-qwen3-5-4b	$0.05	4.4 minmeasured	1.1M+48%	0.60
Frontier reasoning modelllm Highest qualityllm-frontier	answers here	GLM 4.7 Flash200K context · Z.ai via ollama-glm-4-7-flash	$0.15	26 minmeasured	44.5K-71%	0.93
gpt-oss-20b (self-hosted)llm Best value Cheapestgpt-oss	answers here	gpt-oss-20b128K context · OpenAI via ollama-gpt-oss-20b	$0.04	35 minmeasured	26.1K+17%	0.78
gpt-oss-20b (ollama-gpt-oss-20b)llmopenai/gpt-oss-20b@ollama-gpt-oss-20b	answers here	gpt-oss-20b131.1K context · OpenAI via ollama-gpt-oss-20b	$0.04	24 minmeasured	25.2K-12%	0.36
Mid-tier modelllmllm-mid	answers here	gpt-oss-20bOpenAI via ollama-gpt-oss-20b	$0.04	30 minmeasured	19.5K-56%	0.75
Small fast modelllmllm-small	answers here	qwen3.5:4bvia ollama-qwen3-5-4b	$0.20	17 minmeasured	7.7K-59%	0.55
Code modelllmcoder-small	answers here	GLM 4.7 Flash32K context · Z.ai via ollama-glm-4-7-flash	$0.15	42.9 smeasured	2.6K-18%	0.66
Translation skillskillskill-translate	answers here	translategemma:4bvia ollama-translategemma-4b	$0.50	12.7 smeasured	206-7%	0.85
BWeb searchtooltool-web-search	answers here	web-searchvia builtin	free	10.6 msmeasured	47+327%	0.70
On-prem private modelllm Fastestllm-onprem	answers here	qwen3.5:4bvia ollama-qwen3-5-4b	$1.00	14.7 msmeasured	12-100%	0.70
BCalculatortooltool-calculator	answers here	calculatorvia builtin	free	11.2 msmeasured	7new	0.99
UImage generatortooltool-image-gen	decision only	no provider attached	$40	5.3 msmeasured	6new	0.85
UBrowser and desktop automation agentagentautomation-agent	decision only	no provider attached	$5.00	22.0 msmeasured	4+300%	0.78
HHuman escalation queuehumanhuman-escalation	answers here	a person on the queue	$500	22.7 msmeasured	2new	0.90
Coding agentagentcoding-agent	answers here	GLM 4.7 FlashZ.ai via ollama-glm-4-7-flash	$8.00	6.00 sdeclared	0no traffic	0.85
Embedding modelllmembed-small	answers here	qwen3-embedding:4bvia ollama-qwen3-embedding-4b	$0.02	150.0 msdeclared	0no traffic	0.70
UFriendly support personapersonapersona-friendly-support	decision only	the model the router picks	free	-	0no traffic	0.75
ULegal counsel personapersonapersona-legal-counsel	decision only	the model the router picks	free	-	0no traffic	0.80
USenior engineer personapersonapersona-senior-engineer	decision only	the model the router picks	free	-	0no traffic	0.85
Research agentagentresearch-agent	answers here	deepseek-r1:8bvia ollama-deepseek-r1-8b	$6.00	8.00 sdeclared	0no traffic	0.80
USQL skillskillskill-sql	decision only	the model the router picks	$1.00	700.0 msdeclared	0no traffic	0.80
USummarization skillskillskill-summarize	decision only	the model the router picks	$0.80	1.20 sdeclared	0no traffic	0.82
SText to speechtoolspeak-small	answers here	speaches-ai/Kokoro-82M-v1.0-ONNXvia speaches	$60	2.50 sdeclared	0no traffic	0.80
USRE and observability agentagentsre-agent	decision only	no provider attached	$8.00	9.00 sdeclared	0no traffic	0.84
Customer support agentagentsupport-agent	answers here	qwen3.5:4bvia ollama-qwen3-5-4b	$2.00	1.50 sdeclared	0no traffic	0.80
SSpeech to texttooltranscribe-small	answers here	deepdml/faster-whisper-large-v3-turbo-ct2via speaches	free	4.00 sdeclared	0no traffic	0.80
Document and image reader (OCR)llmvision-ocr	answers here	glm-ocr256K context · via ollama-glm-ocr	$0.10	1.50 sdeclared	0no traffic	0.74
UCustomer notification workflowworkflowworkflow-notify-customer	decision only	no provider attached	free	2.00 sdeclared	0no traffic	0.90

The same rows are at /api/v1/models. On your own deployment the targets come from targets.yaml; set metadata.model to a catalogue id such as anthropic/claude-sonnet-4 to link a target to its reference model - see the guide.

A catalogue you do not have to keep

Pick from the whole market, pay for the one model that fits the request.

Every row on this page is a live number: the list price a vendor publishes, the price you pay here by the token, and what the router actually chose for real traffic. Nobody has to maintain a spreadsheet of models - point the client at the router, and when a cheaper model that clears the quality bar appears, the next request goes there.

Photo: Mizuno K

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

Stop choosing a model. Let each request choose.

Point your client at the router with model set to auto; every answer names the model that took it, what it cost and what the alternatives would have cost.

Create a free account Try a request first

Models and routing targets

Recommended per use case

Small writing model

Frontier reasoning model

gpt-oss-20b (self-hosted)

gpt-oss-20b (ollama-gpt-oss-20b)

Mid-tier model

Small fast model

Code model

Translation skill

Web search

On-prem private model

Calculator

Image generator

Browser and desktop automation agent

Human escalation queue

Coding agent

Embedding model

Friendly support persona

Legal counsel persona

Senior engineer persona

Research agent

SQL skill

Summarization skill

Text to speech

SRE and observability agent

Customer support agent

Speech to text

Document and image reader (OCR)

Customer notification workflow

Pick from the whole market, pay for the one model that fits the request.

What it means for which models you pay for.

The right model, not the famous one

Lower cost per request

Prices you can compare

Stop choosing a model. Let each request choose.