Reference

A1.5 run manifest

One place that records exactly what was run: the identifiers, versions, and settings behind every number on this site, and the parts of the record that are incomplete.

Elsewhere the interface uses readable labels (“OpenAI Strong”) because they are easier to compare at a glance. The exact identifiers live here, so nothing is hidden — only kept out of the way.

Systems under test

Label on this siteProviderModel identifierCorrection policy
Qwen 3B + guardrailbasetenQwen/Qwen2.5-3B-Instructranking_guardrail
Qwen 7B + guardrailbasetenQwen/Qwen2.5-7B-Instructranking_guardrail
Claude Haiku 4.5bedrockus.anthropic.claude-haiku-4-5-20251001-v1:0none
Claude Sonnet 4.6bedrockus.anthropic.claude-sonnet-4-6none
OpenAI Smallopenaigpt-5.6-lunanone
OpenAI Strongopenaigpt-5.6-solnone

Self-hosted systems ran on a dedicated Baseten deployment; the Claude systems ran on Amazon Bedrock; the GPT systems ran on the OpenAI API. Deployment handles and endpoint hosts are operational detail and are deliberately not published — the model identifier is what determines the result.

Versions

Runa15-model-comparison-2026-07-24
Dataset[email protected]88 tasks (20 development, 60 holdout, 8 adversarial)
Promptbaseline-intent-prompt@1
Scorer / verifierexact-query-intent@1 — field-for-field equality on pattern_id, metric, dimensions, filters, sort, limit
Correction policyranking_guardrail, a deterministic post-model rule. Applied to the two Qwen systems; the other four ran uncorrected.
Execution dateJul 24, 2026

Decoding and reasoning settings

All three provider paths request a strict JSON schema rather than free text, and cap the response at 256 output tokens. Sampling controls differ because the APIs differ — which is itself part of the configuration:

PathSettings sent
Bedrock (Claude)temperature: 0, maxTokens: 256, json_schema output format
OpenAI (GPT)reasoning_effort: "none", max_completion_tokens: 256, response_format: json_schema (strict). No temperature is sent.
Baseten (Qwen, vLLM)max_tokens: 256, vLLM structured_outputs schema extension. No temperature is sent.
Because the sampling controls are not identical across paths, this run is a comparison of systems as configured, not a controlled experiment isolating the model. Two systems differing here differ in more than their weights.

Execution

Attempts per task

One. No retries, no best-of-n, no self-consistency. Every score is a single sample.

Trials

88 tasks × 6 systems = 528 trials, run sequentially.

Splits

20 development · 60 holdout · 8 adversarial. Only the held-out split is an unbiased generalization test.

Artifacts

Known limitations

Gaps in this record

Two things this manifest cannot currently pin down, recorded here rather than papered over.Closing both is the next reproducibility step, ahead of repeated trials.