Ramp Labs Says 35B Qwen 3.5 Model Outperforms Opus 4.6 on Financial Question-Answering

Ramp Labs said its 35 billion parameter Qwen 3.5 retrieval model beat Opus 4.6 on exact-task accuracy for financial question-answering while running at far lower latency. StartupHub.ai reported the demo on November 5, 2026, ahead of the November launch of Ramp Sheets.

Ramp Labs' Post-Training Bet With Prime Intellect

Ramp Labs paired with Prime Intellect to post-train the 35 billion parameter Qwen 3.5 retrieval model, Alexander Shevchenko said. The smaller system beat Opus 4.6 on exact-task accuracy for financial question-answering while running at far lower latency. Shevchenko said the team could not wait for the next frontier jump: "We need to serve our financial professionals now."

Prime Intellect's Will Brown argued on November 2, 2026, that companies should train their own models for about $300,000 and update weights nightly, rather than re-prompting frontier systems, StartupHub.ai reported.

Ramp Labs also demoed 10,000 parallel agents that pen-tested Ramp's own platform and found 7 high-severity vulnerabilities reachable by remote email. Shevchenko said one chain required stitching multiple safe components to achieve Lua script injection via email, and that open-weight models at five times lower cost delivered near-frontier detection when run as a swarm. The seven issues were patched, he said.

A second technique, KV cache cartridges, compresses 20 to 30 times more context into the cache offline, Shevchenko said, so month-end close books can be prepared overnight. The agent opens on the first of the month with ledgers, aging reports and documents already loaded, with no agentic retrieval needed.

Ramp's Funding and AI Unit Records

According to akta.pro company records, Ramp raised a $750 million Series F in June 2026 at a $44 billion post-money valuation, led by ICONIQ, GIC and Ontario Teachers' Pension Plan. The same records list a $1.9 billion Series G in October 2026 led by Avenir, Dragoneer Investment Group and Founders Fund, and total funding of $5.9 billion at Series G stage.

The records also show Ramp launched Ramp Applied AI Solutions in June 2026, a service unit that embeds engineers within customer organizations to build bespoke AI systems on Ramp's platform for large enterprise finance teams.

Ramp's Earlier AI Spend Data

On November 4, 2026, StartupHub.ai reported Ramp data showing AI spending per employee averages $12 monthly, with the top 1% near $10,000. Firms spending $30 per employee per month hired 12% more than non-adopters, and the top AI quartile roughly doubled revenue over three years.

On November 5, 2026, StartupHub.ai reported Ramp data showing AI bills are halved for the heaviest users due to price cuts, according to Ramp's economist.

Ramp Router Strategies Ahead

Shevchenko said more routing strategies for Ramp Router are scheduled to launch later, alongside additional Ramp Labs posts on Portal and self-improving monitors.