We Raised Your AI Credit Limits Across Every Plan
We are raising the AI Credit limits across every plan. The Standard plan goes from 100 to 2,000 credits per day. Pro goes from 1,000 to 10,000. The free evaluation stops being a 5-per-day trickle and becomes a 1,500-credit allowance you can spend whenever you want.
We can do this because our agents now run on far fewer tokens, and without giving up any quality. There is more on how below. First, what you get:
| Plan | Before | Now |
|---|---|---|
| Standard | 100 credits per day | 2,000 per day and 4,000 per week |
| Pro | 1,000 credits per day | 10,000 per day and 20,000 per week |
| Evaluation | 5 credits per day | 1,500 credits, one-time |
Standard and Pro carry both a daily and a weekly cap. Both apply at once, so whichever you reach first is the one that stops you.
Based on our evaluation data, a chat turn is worth around 10 AI Credits. The Standard plan now covers about 200 chat turns a day, against 10 before. Deep Research Agents vary more in what they consume, but you will be able to run many more of them before reaching your limits.
We are also changing how the Evaluation plan works, based on client feedback. Five credits a day was enough to see the interface but not enough to test anything. It is now a one-time 1,500-credit allowance that lets you run a real analysis on a company you cover, at whatever pace suits you, instead of coming back over several days to do a proper evaluation.
The new limits are already live. They applied automatically, with nothing to enable and no plan change required. Open AI Analyst Chat or start a Deep Research Agent, or check the pricing page to see what each plan now includes.
What changed on our side
We could raise the limits this far because each request now costs us far less to run. Over the last few months we rebuilt the harness the agents run inside: tighter tool definitions, less redundant context on each model call, fewer wasted round-trips between an agent and the documents it reads, and caching so we stop paying twice for the same report state.
Cost was not the only thing that improved. The same work made the AI meaningfully better. The agents now reason over cleaner context, spend fewer steps on plumbing and more on the analysis, and return answers that are more accurate and better grounded in the source documents. The product got stronger and cheaper for us to run at the same time, and we would rather hand that saving back than book it.
Frequently asked questions
Alex is the co-founder and CEO of Marvin Labs. Prior to that, he spent five years in credit structuring and investments at Credit Suisse. He also spent six years as co-founder and CTO at TNX Logistics, which exited via a trade sale. In addition, Alex spent three years in special-situation investments at SIG-i Capital.
