---
title: "AfterQuery's agent training data lifted OpenAI's small model from 3.1% to 17% on Terminal-Bench"
description: "The November 4 run, showcased on Thinking Machines' Tinker, sits beside rival Mercor's proof on the same page, giving labs a direct comparison of measured benchmark uplift for the first time. It is the claim beneath AfterQuery's reported $3.2 billion valuation, now competing in public for labs' data budgets."
publisher: "The Inference"
section: "Model"
published: 2026-11-09T15:38:28.729Z
modified: 2026-11-09T15:38:28.729Z
canonical: https://theinference.org/article/afterquery-s-agent-training-data-lifted-openai-s-small-model-from-3-1-to-17-on-terminal-bench
language: en
keywords: "AI Training Data, Benchmarks, Agents, Mercor, Thinking Machines Lab, OpenAI"
---

# AfterQuery's agent training data lifted OpenAI's small model from 3.1% to 17% on Terminal-Bench

> The November 4 run, showcased on Thinking Machines' Tinker, sits beside rival Mercor's proof on the same page, giving labs a direct comparison of measured benchmark uplift for the first time. It is the claim beneath AfterQuery's reported $3.2 billion valuation, now competing in public for labs' data budgets.

On November 4, 2026, [AfterQuery](https://theinference.org/markets/companies/afterquery) published the kind of result AI labs normally see only in sales meetings [1][2]. It post-trained gpt-oss-20b, [OpenAI](https://theinference.org/markets/companies/openai)'s small open-weight model, on its terminal-agent data, and the model's pass rate on Terminal-Bench 2.0, a benchmark of real command-line work, went from 3.1% to 17.0% [1][2]. The recipe: supervised fine-tuning on successful agent trajectories, then reinforcement learning in Harbor sandbox environments that reward finished work [2].

The host matters as much as the number. [Thinking Machines Lab](https://theinference.org/markets/companies/thinking-machines-lab), founded by former OpenAI CTO Mira Murati, put the run on [the customer showcase of Tinker](https://thinkingmachines.ai/tinker/), its training service, where [AfterQuery](https://x.com/AfterQuery/status/2118039702163112106)'s biggest rival appears beside it [2]. The company has [published uplift numbers on its own blog before](https://www.afterquery.com/) [6]. This one is on neutral ground.

## Validation stopped being the lab's job

Tinker rents the training loop itself: open-weight models from a few billion to a trillion parameters, adapted with LoRA, billed by the token [2]. That is the hop that changed the data business. A vendor with no compute of its own can now do what a lab researcher does, train a model on its data and measure what moves. AfterQuery co-founder Spencer Mateega told [Forbes](https://www.forbes.com/sites/annatong/2026/09/01/afterquery-becomes-ycs-fastest-unicorn-at-32-billion/) in April that this internal pipeline is how the company proves quality before labs ever see the data [4]. This month the proof went public.

## The rival is on the same page

[Mercor](https://theinference.org/markets/companies/mercor), the larger rival, trained [Nvidia](https://theinference.org/markets/companies/nvidia)'s Nemotron 3 Ultra on more than a thousand banking, law and consulting tasks and doubled the model's first-attempt success on the held-out APEX-Agents benchmark [2]. Two vendors, one claim, measured the same way, on a page any buyer can score.

*Chart: **Two data vendors, one showcase page: both publish a before-and-after gain** Range chart: AfterQuery's gpt-oss-20b pass rate rising from 3.1% to 17.0% on Terminal-Bench 2.0, and Mercor's Nemotron 3 Ultra rising from 11.5% to 22.9% on APEX-Agents.*

*Pass rates before and after training on each vendor's data, on two different benchmarks. AfterQuery: Terminal-Bench 2.0 pass@1 averaged over five runs on 89 tasks. Mercor: first-attempt success on 480 held-out APEX-Agents tasks under identical evaluation conditions. Source: Thinking Machines Lab, Tinker customer showcase. [2]*

## The proof is about data, not a deployed agent

Read the result honestly: a 17% pass rate is evidence about data quality, not a tool anyone can put to work, and the strongest agents now clear roughly two thirds of the tasks on [Terminal-Bench's current version](https://www.tbench.ai/) [3]. The open link is transfer, whether data that lifts a small open model also lifts a frontier training run. No vendor has published that, and the first to show it sets the market's price.

## The money is already positioned

The round Forbes reported on September 1 values AfterQuery at $3.2 billion; it is reportedly profitable, with revenue its co-founder puts in the hundreds of millions [4][5]. Mercor is reportedly in talks with Nvidia at a $20 billion valuation [4]. Nvidia sits on the demand side twice: its Nemotron models were trained with AfterQuery data, and it may bankroll Mercor [4][5]. Behind both vendors are [paid experts recording how they actually work](https://theinference.org/article/mercor-s-career-of-the-future-pays-95-an-hour-and-the-hours-are-in-dispute), and their judgment is the scarce input [6].

The vendors that measure what they sell deserve those budgets, because the measuring is what turns expertise into something a lab can buy. The next number to watch is an uplift attached to a frontier model. Until it appears, anyone can keep score.

## Takeaway

AfterQuery's benchmark proof now sits in public beside Mercor's on the same page, and labs can score data vendors on measured uplift before they spend.

## Sources

1. [AfterQuery, post on X, 4 November 2026](https://x.com/AfterQuery/status/2118039702163112106)
2. [Thinking Machines Lab, Tinker customer showcase, accessed 9 November 2026](https://thinkingmachines.ai/tinker/)
3. [Terminal-Bench leaderboard, accessed 9 November 2026](https://www.tbench.ai/)
4. [Forbes, Anna Tong, "AfterQuery Becomes YC's Fastest Unicorn At $3.2 Billion", 1 September 2026](https://www.forbes.com/sites/annatong/2026/09/01/afterquery-becomes-ycs-fastest-unicorn-at-32-billion/)
5. [TechCrunch, Julie Bort, "AfterQuery reportedly becomes Y Combinator's fastest-ever unicorn, now valued at $3.2B", 1 September 2026](https://techcrunch.com/2026/09/01/afterquery-reportedly-becomes-y-combinators-fastest-ever-unicorn-now-valued-at-3-2b/)
6. [AfterQuery, company website and research blog, accessed 9 November 2026](https://www.afterquery.com/)


---

AfterQuery's agent training data lifted OpenAI's small model from 3.1% to 17% on Terminal-Bench — The Inference. Canonical: https://theinference.org/article/afterquery-s-agent-training-data-lifted-openai-s-small-model-from-3-1-to-17-on-terminal-bench