The newest frontier models are live in datavessel. Claude Opus 5 — Anthropic’s new flagship — and OpenAI’s GPT-5.6 family (Sol, Terra, and Luna) are now available to every agent on the platform. Open the AI Keys section, pick the model you want per provider, and every agent you run — order desk, dispute evidence, win-back, custom — starts thinking with it immediately.
Because datavessel is bring-your-own-key, there’s nothing to wait for on our side and no markup in the middle: when a provider ships a better brain, you plug it straight into your workforce.
Claude Opus 5 — the new Anthropic flagship
Claude Opus 5 (claude-opus-5) is now the top of the Anthropic list in datavessel, and the default for new Anthropic keys. It’s vision-capable and built for exactly the kind of work our agents do all day: long multi-step tool use, careful reasoning over your store’s real records, and knowing the difference between what the data shows and what it doesn’t.
If you’ve already chosen a model, nothing changes without you — your selection stays exactly as you set it. New keys and unselected users get Opus 5 by default.
GPT-5.6 — three tiers: Sol, Terra, Luna
OpenAI’s GPT-5.6 generation arrives as a family, and we’ve added all three:
- GPT-5.6 Sol (
gpt-5.6-sol) — the frontier tier, and the new default for OpenAI keys. Maximum reasoning quality for judgment-heavy work. - GPT-5.6 Terra (
gpt-5.6-terra) — the balanced tier: strong reasoning at roughly half the cost of Sol. - GPT-5.6 Luna (
gpt-5.6-luna) — the cost-efficient tier, ideal for high-volume routine runs.
All three take text and images and handle the long-context work agents actually need — big order queues, long tool transcripts, full campaign histories in a single pass. The family is available whether your key is with OpenAI directly or Azure OpenAI; both providers get all three tiers.
Match the model to the job
Three tiers isn’t marketing — it’s how you should run a workforce. Not every agent needs the frontier model:
- Put Opus 5 or Sol on the judgment-heavy jobs: the Dispute Evidence Desk building a representment file, or a custom agent making merchandising calls.
- Run Terra for the everyday middle: order-desk sweeps, report writing, campaign recaps.
- Let Luna handle the high-frequency routine: scheduled checks and simple classify-and-file work where speed and cost win.
Because you choose per provider key in AI Keys, switching tiers is a dropdown, not a migration.
How to switch
In the datavessel app, open AI Keys. Each connected provider — Anthropic, OpenAI, Azure OpenAI — shows its model list with the new ones at the top. Pick one, and it takes effect for your agents right away. That’s the whole process.
Three things to know:
- Your existing choice is never overridden. Defaults changed for new and unselected keys only.
- BYOK means provider rates. You pay Anthropic or OpenAI directly at their prices — datavessel adds no per-token margin.
- Every surface follows your choice — the app, scheduled agent runs, the datavessel CLI, and Claude via MCP all use the model you set.
Why this matters for autopilot
An agent platform is only as good as the model reasoning inside it — and model generations are turning over fast. Our job is to make sure the day a better model exists, your order desk, your dispute evidence, and your marketing agents can run on it. Today that’s Claude Opus 5 and GPT-5.6. Connect your key, pick your tier, and let the workforce get smarter.
You bring the key. The agents bring the work.
datavessel is e-commerce autopilot for Shopify, Shopware, and WooCommerce — analytics, SEO, store ops, order desk, and dispute evidence, running on the model you choose. Learn what e-commerce autopilot is →
Leave a Reply