Three New Models Land in BotTrottr: Grok 4.7, Gemini 3.8 Flash & Gemini 3.1 Pro
Every agent in BotTrottr runs on a model, and you pick which one. As of today there are three new options in that list: Grok 4.7, Gemini 3.8 Flash, and Gemini 3.1 Pro.
They’re available right now in the model picker — when you deploy a new agent, when you edit an existing one, and on the agent cards in your dashboard.
Pick the right tool, not the “best” model
There’s no single winning model. There’s a model that’s right for the job your agent does every hour, and a model that’s right for the job it does once a week and has to get exactly right.
These three cover a genuinely wide range of that spectrum — from cheap-and-fast enough to run on a tight loop, to frontier-tier reasoning that costs real money per run and earns it.
Grok 4.7 — frontier reasoning with a live view of the world
Grok 4.7 is xAI’s newest frontier model, built for coding, agentic work, and knowledge tasks. It carries a 500,000-token context window and a knowledge cutoff in May 2026.
The part that matters most for agents: Grok models have native, real-time web search built in. For an agent whose entire job is “tell me what changed” — a news beat, a competitor’s pricing page, a market that moves overnight — that’s the difference between a model that can go look and one that can only recall. BotTrottr routes Grok’s search natively rather than bolting on a separate search step.
It’s also cheaper than the Grok 4 it sits above — $2.00 per million input tokens and $6.00 per million output, against Grok 4’s $5.00/$20.00. Newer, more capable, and less than half the price on input. If you were running agents on Grok 4, this is a straightforward upgrade.
Best for: research and monitoring agents, competitive and market digests, anything where “as of right now” matters more than “as of training.”
Gemini 3.8 Flash — the fast, cheap workhorse
Gemini 3.8 Flash is Google’s latest in the Flash line, and it’s the one to reach for when an agent runs often and the per-run cost actually shows up on your bill.
At $0.75 per million input tokens and $3.75 per million output, it’s roughly a third the input cost of Gemini 3.1 Pro and several times cheaper than the frontier-tier models — and it’s currently even cheaper on Google’s promotional pricing through the end of 2026.
Cheap doesn’t mean limited here. Flash-class models have become genuinely good at the tasks most scheduled agents actually do: summarise this, extract the key points, classify these items, draft a reply, reformat a feed into something readable. Those are high-volume, low-drama jobs. Paying frontier prices for them is the most common way to waste money on automation.
If your agent runs hourly and you’re on your own API key, start here. Move up only when you can point at a specific run where Flash wasn’t good enough.
Best for: hourly and daily digests, inbox triage, summarisation, classification, high-volume “just tell me what’s important” work.
Gemini 3.1 Pro — when the answer has to be right
Gemini 3.1 Pro is the deep-reasoning option: $2.00 per million input tokens and $12.00 per million output, with a 200k-token context before long-context pricing kicks in.
That’s a real step up in cost, and it buys a real step up in reasoning depth — the kind that shows on multi-step analysis, long-document synthesis, and tasks where being subtly wrong is worse than being slow. An agent that reads a contract, reconciles a set of numbers, or works through a chain of dependent conclusions is exactly where the extra spend pays for itself.
The honest framing: Pro-tier models are a waste on a summarise-and-email agent, and a bargain on an agent whose output you’re going to act on. Match the spend to the stakes.
Best for: deep analysis, long-document work, multi-step reasoning, anything where a wrong answer costs more than the run did.
A quick way to decide
- “It runs every hour and I just need it summarised.” → Gemini 3.8 Flash.
- “It needs to know what happened today.” → Grok 4.7.
- “If this is wrong, it’s a problem.” → Gemini 3.1 Pro.
And you don’t have to pick one for the whole account. Models are per-agent, so an expensive reasoning agent and a cheap high-frequency agent can run side by side, each on the model that earns its keep.
What changed in the picker
We also retired Gemini 2.5 Flash-Lite — the model you may have seen marked as old. If you had an agent on it, switch it to Gemini 3.8 Flash: same fast-and-cheap role in the lineup, a generation newer, and better at the work. Everything else about that agent stays exactly the same.
Start using them
Open any agent — new or existing — and you’ll find all three in the model list. Nothing else to enable.
- New to BotTrottr? Create an account and deploy your first agent in a couple of minutes.
- Already running agents? Edit one, switch its model, and hit Run to see the difference on a live task before you commit to it.
Not sure how to think about building the agent itself? Start with What is BotTrottr? for the lay of the land, then Teams, Swarms & Skills for how to organise more than one.
Follow along on the RSS feed — we’ll keep adding models as they land.