OSWorld Windows tasks on Daytona Windows sandboxes C

AI benchmark · Coding

Tier
C why
Snapshot

Open benchmark ↗daytona.io/dotfiles/snapshots-as-search-states-go-explore-wi…

Tier C

C

Niche: a handful of models scored, or built by an RL vendor.

How much stands behind a tier: usage, who backs it, a confirmed line from its own site or paper, and a clear read on X. High with three of the four, medium with two, low otherwise.

  • listed with Daytona
  • no proper name (a paper title or description)
  • in curated RL directories (2 pts)
How we got to tier C
  1. Adoption: listed with Daytona; no proper name (a paper title or description). Tier D.
  2. Not yet read on X.
  3. In curated RL directories: 2 points, a floor of C. Raised to C.
  4. Not yet confirmed by a line from its own site or paper.

A clear view on X moves a tier one step up or down; most entries don't move. How tiers work.

Training methods

Methods OSWorld Windows tasks on Daytona Windows sandboxes uses, offers or builds for, with public evidence: its own website says so, or two independent sites do.

  • “This guide presents a simple, self-contained script that performs reinforcement learning training of Qwen3-1.7B-Base . In particular, we use reinforcement learning with verifiable rewards , with the reward being computed from the test pass rate of model-written functions.”

    daytona.io
  • “…pt per step) 2. Evaluate : Each completion runs in its own Daytona sandbox against a test suite 3. Reward : Completions that pass more tests get higher rewards; errors or banned patterns get negative rewards 4. Update : GRPO reinforces completions that scored above their group average”

    daytona.io

    “- run.py — A single-episode demo that creates one sandbox, runs one complete interaction, and tears down. Useful for understanding the environment and verifying your setup. - train.py — A full GRPO training loop that creates 500 sandboxes, collects multi-turn rollouts in parallel with batched vLLM generation, runs policy gradient updates with LoRA, and hot-swaps adapters into vLLM between iterations.”

    daytona.io

Not on the map yet: no line from its own site or paper places it in a domain.

Next: browse all benchmarks. Entry from the RL Research daily scrape of public sources, 2026-11-16 snapshot.