Vantage RuntimeAI · Pricing
One continuous trust decision on one critical path — go/no-go across functionality, cost, reliability, safety, and compliance. Free vantage-core decides on your machine or CI. Paid protects history, baselines, and review — not permission to decide.
Free forever
$0
pip install vantage-core — no RuntimeAI account (-U if upgrading)vantage-core demo — Acme sample / offline talk-trackvantage-core init — samples + starter YAML contractssuite rerun --baseline latest · CI stub · portable decision · exit 0/2/1When it has to survive past merge
Pro $49/mo
Preflight sizes one path before you spend (try a prefilled run → — then pin the same contract in free CI.
Early pricing: Pro is $49/mo per shipped product — the land price for still-trust and retained history on a critical path. Team & Enterprise are per account, later — hello@vantageai.cc.
Same path, five models — sorted by score, then price. Illustrative rulers you can re-run in CI, not the paid product. Each row links to the scorecard PDF.
Which model audits a buyer-facing page for ship-gate clarity — keep/cut/rewrite — without overpaying?
| Model | Score | Pass | Est. $/run | Report |
|---|---|---|---|---|
claude-haiku-4.5 | 7.6 | Pass | $0.042 | Report · PDF |
nova-lite-v1 | 7.2 | Pass | $0.0009 | Report · PDF |
claude-sonnet-4.6 | 7.2 | Pass | $0.049 | Report · PDF |
kimi-k2 | 3.2 | Fail | $0.0083 | Report · PDF |
mistral-nemo | 2.8 | Fail | $0.0002 | Report · PDF |
Spread 7.6 → 2.8. Cheapest decision pass: nova-lite at ~$0.0009/run. Haiku leads at 7.6; micro models and kimi fail the bar.
Which model catches train/serve leakage before a bad feature ships?
| Model | Score | Pass | Est. $/run | Report |
|---|---|---|---|---|
nova-micro-v1 | 8.0 | Pass | $0.0006 | Report · PDF |
gpt-4.1-mini | 8.0 | Pass | $0.0070 | Report · PDF |
gpt-oss-20b | 7.2 | Pass | $0.0006 | Report · PDF |
gemini-2.5-flash | 7.2 | Pass | $0.0080 | Report · PDF |
mistral-nemo | 6.8 | Fail | $0.0002 | Report · PDF |
Scores split 8.0 / 7.2 / 6.8. Cheapest full pass: nova-micro at ~$0.0006/run. Mistral fails the bar; gpt-oss and Gemini clear it at 7.2.
Which model drafts outreach that clears a fixed pass bar — without overpaying?
| Model | Score | Pass | Est. $/run | Report |
|---|---|---|---|---|
nova-lite-v1 | 7.2 | Pass | $0.0043 | Report · PDF |
claude-haiku-4.5 | 7.2 | Pass | $0.081 | Report · PDF |
mistral-nemo | 6.8 | Fail | $0.0007 | Report · PDF |
gpt-4o-mini | 6.4 | Fail | $0.0074 | Report · PDF |
llama-3.1-8b-instruct | 5.6 | Fail | $0.0007 | Report · PDF |
Wide spread: 7.2 down to 5.6. Two models pass (nova-lite and Haiku); three fail. Cheapest pass is nova-lite at ~$0.004/run — ~19× cheaper than Haiku for the same score.
Different jobs, different units — public list prices for orientation only.
| RuntimeAI | Braintrust | LangSmith | Promptfoo | |
|---|---|---|---|---|
| Core job | Continuous ship / still-trust (five axes) | Experiments + LLM-judge evals | Traces + debug + evals | Local assertions / red-team |
| Free surface | vantage-core CI gate |
Starter tier | Developer tier | OSS / local CI |
| Paid from | Pro $49/mo | Pro ~$249/mo flat | Plus ~$39/seat/mo | Enterprise sales |
| Paid unit | Protected suite / release evidence | Processed data + scores | Traces + seats | Support / hosted |
| You still pay | Your stack + inference | Your stack + inference | Your stack + inference | Your stack + inference |
Reference figures are approximate public list prices (mid-2026). See FAQ for product-fit detail.
Vantage prices the continuous trust decision beside your stack — not a model vendor, not orchestration, not traces. Go/no-go across functionality, cost, reliability, safety, and compliance. Free: pip install vantage-core, local contracts, runtimeai.decision/v1, CI exit, BYOK OpenRouter. Paid: Pro $49/mo per shipped product for baselines, history, still-trust re-checks, release evidence, and model watch when ready. Team and Enterprise are per account, later. Preflight is an on-ramp only.