ModelSpec is an analysis of alternatives for AI models: it decides which models fit a job from your requirements, sourced benchmarks and real cost, and shows its work.

Regulated data, best available

Find the strongest assistant among offerings with the data-handling commitments regulated workloads need.

Worked example as of 2026-09-23, from the public data image.

HIPAA-ready offerings; strongest assistant, cost no object.

Requirements

offering.data.trains_on_customer_data = No: Customer prompts and outputs must not be available for model training.
offering.data.zero_retention = Yes: The provider must offer a mode that stores neither prompts nor outputs.
offering.attestation.baa = Yes: The provider must offer a HIPAA business associate agreement for the service.

Preferences

chat_preference: Rank only by assistant quality among compliant offerings.

Applied template spec

spec_version

where

offering.data.trains_on_customer_data = No
offering.data.zero_retention = Yes
offering.attestation.baa = Yes

optimize

weights

chat_preference: 1.0

Worked answer

Decision status: answered.

Separated on the evidence

anthropic/claude-opus-5-5

probability bands: a model joins the leader when P(its score >= the leader's) >= 0.25 under the score posteriors; a weighted capability with an 80% interval wider than 2.8 is not enough evidence

Score and score interval use the objective's dimensionless normalised units. Cost per task is for the engine-selected offering and the applied spec's task size. When task size is omitted, the engine uses its documented default. P(best) is the engine probability. Unavailable means the engine did not return a value. Thin evidence is undetermined and is not a recommendation.

Best band

Model	Score	Score interval	P(best)	Cost per task, USD	Capability estimates and intervals
`anthropic/claude-opus-5-5`	1.0	[0.451151, 1.548849]	0.75	0.24	chat_preference: 1.584232494; interval [0.565758585, 2.602706403]; benchmarks 1; direct benchmarks 1

Rest band

Model	Score	Score interval	P(best)	Cost per task, USD	Capability estimates and intervals
`anthropic/claude-fable-5`	0.704386	[0.624333, 0.784439]	0.23046875	0.6	chat_preference: 1.035674964; interval [0.887124662, 1.184225266]; benchmarks 34; direct benchmarks 34
`anthropic/claude-opus-4-6`	0.564081	[0.486275, 0.641887]	0.00390625	0.3	chat_preference: 0.775317472; interval [0.630936802, 0.919698141]; benchmarks 36; direct benchmarks 36
`anthropic/claude-fable-5-1`	0.563792	[0.479971, 0.647613]	0.015625	0.6	chat_preference: 0.774780684; interval [0.619237933, 0.930323435]; benchmarks 31; direct benchmarks 31
`anthropic/claude-opus-4-7`	0.505909	[0.425856, 0.585962]	0.0	0.3	chat_preference: 0.667370368; interval [0.518820066, 0.81592067]; benchmarks 34; direct benchmarks 34
`anthropic/claude-opus-5`	0.427813	[0.34776, 0.507865]	0.0	0.3	chat_preference: 0.522450073; interval [0.373899771, 0.671000375]; benchmarks 34; direct benchmarks 34
`google/gemini-3-8-flash`	0.396869	[0.301665, 0.492073]	0.0	0.045	chat_preference: 0.465029132; interval [0.288363151, 0.641695113]; benchmarks 24; direct benchmarks 24
`google/gemini-3-7-flash`	0.362924	[0.26772, 0.458128]	0.0	0.045	chat_preference: 0.402038706; interval [0.225372725, 0.578704687]; benchmarks 24; direct benchmarks 24
`google/gemini-3-1-pro-preview`	0.152223	[0.07217, 0.232276]	0.0	0.128	chat_preference: 0.01105026; interval [-0.137500042, 0.159600562]; benchmarks 34; direct benchmarks 34
`google/gemini-3-5-flash`	0.0	[-0.080053, 0.080053]	0.0	0.096	chat_preference: -0.271422803; interval [-0.419973105, -0.122872501]; benchmarks 34; direct benchmarks 34

Engine warnings

not_separable

Decision dec_deeb0e6f23ab984de448de39 on public snapshot snap_05a963d2477ac2fb.

What this template teaches

Governance promises are Musts; with no cost term, quality alone ranks what passes.

The live board uses current data. Its answer can change as evidence changes.

Apply this template on Decide

Human lookup is free on Decide. CLI, API, and MCP access requires a key. Access and pricing.

Use-case examples · How we decide