RL vendor landscape

Datacurve sells expert-built RL environments, long-horizon tasks, agent trajectories, SFT demonstrations, off-the-shelf datasets and benchmarks to AI labs, with a coding focus.

Why it's in Coding

• Coding leads with“IntroducingDeepSWE, our long-horizon coding benchmark” source ↗
• Security“Software EngineeringData ScienceCyber SecurityMachine LearningResearch” source ↗
• Enterprise work“enterprise software systems with evolving business logic and real integration constraints” source ↗

What it sells

• RL environments“Durable environments for measuring agentic capabilities where real work happens: long horizons, natural instructions, realistic tools, domain-specific judgment, and partial progress.” source ↗
• Datasets“Prebuilt datasets, curated for signal, reviewed for quality, and structured to drop into your training stack without translation work.” source ↗
• Evals“We build benchmarks that capture task-faithful, domain-sensitive lift” source ↗
• Human data“Demonstrations that set the right behavioral prior for models, captured through bespoke tooling that lets experts work naturally” source ↗

See it in the vendor universe →

Datacurve on RL Research → Website ↗

Bespoke Labs builds company-scale RL environments and infrastructure, plus agent evaluation and prompt/policy optimization (GEPA), for frontier labs and enterprises training long-horizon agents.

Why it's in Coding

• Coding leads with“Company-scale simulations with real codebases and microservices, allowing agents to master the complex, long-horizon workflows required for production.” source ↗
• Enterprise work“Evaluate agents in environments that mirror your systems and processes.” source ↗

What it sells

• RL environments“Bespoke Labs builds company-scale RL environments and infrastructure to let frontier labs and enterprises train long-horizon agents for production.” source ↗
• Evals“Evaluate agents in environments that mirror your systems and processes.” source ↗
• Enterprise AI services“Enterprises rely on us to make AI agents dependable.” source ↗

See it in the vendor universe →

Bespoke Labs on RL Research → Website ↗

Emulated builds ready-to-use post-training datasets and challenging environments for software engineering, ML engineering and autonomous AI research, plus benchmarks.

Why it's in Coding

• Coding leads with“building post-training datasets for software engineering and AI R&D” source ↗
• Science“Investigating alignment and interpretability questions” source ↗

What it sells

• Datasets“We provide ready-to-use datasets across software engineering, machine learning engineering, autonomous research, and model post-training.” source ↗
• RL environments“For models to keep improving, they need more challenging environments” source ↗
• Evals“Software Development Automation Benchmark” source ↗

See it in the vendor universe →

Emulated on RL Research → Website ↗

Abundant builds RL environments and datasets for AI labs and enterprises, focused on domains like coding and physical sciences.

Why it's in Coding

• Coding leads with“Deep in domains like coding, physical sciences, RSI, and alignment” source ↗
• Science“Deep in domains like coding, physical sciences, RSI, and alignment” source ↗

What it sells

• RL environments“We specialize in creating environments and datasets for RL by leveraging our experience in simulation and model training.” source ↗
• Datasets“We specialize in creating environments and datasets for RL by leveraging our experience in simulation and model training.” source ↗

See it in the vendor universe →

Abundant on RL Research → Website ↗

Idler sells reinforcement learning environments and bespoke evaluations built from real production work for frontier AI labs.

Why it's in Coding

• Coding leads with“Frontier Coding” source ↗
• Enterprise work“Economically Valuable Workflows” source ↗

What it sells

• RL environments“Reinforcement learning environments built from real production work” source ↗
• Evals“Bespoke Evaluations” source ↗

See it in the vendor universe →

Idler on RL Research → Website ↗

Mechanize builds high-fidelity RL environments and evals for frontier coding agents, where models do software engineering work scored by a grader.

Why it's in Coding

• Coding leads with“models carry out software engineering work such as building a feature, deploying an application, or debugging an issue in an unfamiliar codebase.” source ↗

What it sells

• RL environments“We build environments and evals for frontier coding agents.” source ↗
• Evals“GBA Eval is a benchmark that measures how well coding agents can write a Game Boy Advance emulator from scratch in 24 hours.” source ↗

See it in the vendor universe →

Mechanize on RL Research → Website ↗

Parsewave sells custom, fully human-authored datasets, solution traces and expert evaluations, mainly real-world software engineering tasks, for training and benchmarking frontier AI models.

Why it's in Coding

• Coding leads with“The most effective approach is to use curated, real-world engineering tasks that represent a reliable ground truth.” source ↗

What it sells

• Human data“Every datapoint is created from scratch by senior practitioners, primarily senior software engineers alongside vetted domain specialists.” source ↗
• Evals“Parsewave datasets are designed for use in supervised fine-tuning, reinforcement learning from human feedback, or as benchmarks to assess model performance.” source ↗

See it in the vendor universe →

Parsewave on RL Research → Website ↗

pre.dev sells a general-purpose coding agent product and, separately, long-horizon RL coding tasks built from licensed private codebases, each shipped as a runnable environment with graded verifiers and solver trajectories.

Why it's in Coding

• Coding leads with“Hard, verified coding tasks. From code no model has seen.” source ↗

What it sells

• RL environments expanding“Long-horizon RL tasks for post-training. Tasks, not repo access: each ships in a runnable environment with a graded verifier” source ↗
• Datasets expanding“Exported as a tar.gz with its sha256, QC evidence and solver trajectories included.” source ↗

See it in the vendor universe →

pre.dev on RL Research → Website ↗

Proximal is a data research lab that builds long-horizon coding RL environments, training data and benchmarks such as FrontierSWE for frontier AI labs.

Why it's in Coding

• Coding leads with“Our ultra-long-horizon coding benchmark. Frontier models clear only a fraction of its tasks.” source ↗
• Security“Our investigation into the CyberGym benchmark and reliable cyber evaluations.” source ↗

What it sells

• Evals“34 tasks, 13 frontier models, 20-hour budgets. Our hardest coding benchmark yet.” source ↗
• RL environments“How we run agentic RL across Modal GPUs and our own Kubernetes sandbox fleet.” source ↗

See it in the vendor universe →

Proximal on RL Research → Website ↗

Quesma sells an analytics product that turns Claude Code, Codex and Cursor session logs into views of what coding agents did, what it cost and what it delivered, and also publishes coding-agent benchmarks and has built training environments for frontier labs.

Why it's in Coding

• Coding leads with“Quesma contributed to Terminal-Bench 3.0, a benchmark for testing agents on practical terminal tasks.” source ↗
• Security“Can coding agents find backdoors in compiled binaries using Ghidra and radare2?” source ↗

What it sells

• RL environments expanding“We built training environments for difficult, multi-hour agent tasks.” source ↗
• Evals expanding“Can coding agents find backdoors in compiled binaries using Ghidra and radare2?” source ↗

See it in the vendor universe →

Quesma on RL Research → Website ↗

Handshake, the student and early-career jobs network, sells expert human data for AI training and publishes research on graders, verifiers and coding-agent benchmarks.

Why it's in Coding

• Coding leads with“Coding Agents Build for the Grader They Imagine, Not the User: Speculative Reward Hacking in DeepSWE” source ↗

What it sells

• Human data“Exploring why human judgment, skill, and expertise remain essential to training modern AI and how people shape the future of artificial intelligence.” source ↗

See it in the vendor universe →

Handshake on RL Research → Website ↗

Veris AI sells a simulation platform that builds high-fidelity twins of an enterprise's systems, data, APIs and users so AI agents can be benchmarked, trained and built against before production.

Why it's in Enterprise work

• Enterprise work leads with“More than 70 are ready today: CRMs, billing, ticketing, databases, queues, clouds.” source ↗
• Coding“A stateful sandbox where coding agents verify their work” source ↗
• Media“VAmoS Bench · 17 voice-agent systems on the same simulated calls” source ↗
• Finance“FX derivatives desk” source ↗

What it sells

• RL environments“A simulated copy of your systems, data, APIs, and users, where AI gets tested, trained, and built against before it ever touches production.” source ↗
• Evals“Build a benchmark around your own policies, workflows, and systems, then grade every agent you buy or build against it, end to end, in a twin of production” source ↗
• Sandboxes and compute expanding“Give coding agents a stateful simulation of your dependencies, so integration tests run and issues get fixed before the PR opens” source ↗

See it in the vendor universe →

Veris AI on RL Research → Website ↗

Andon Labs builds long-horizon agent evals such as Vending-Bench and runs real AI-operated businesses as safety testbeds, and is launching Pion, a platform for agents that run businesses.

Why it's in Enterprise work

• Enterprise work leads with“Can AI agents run vending machine businesses?” source ↗
• Coding“Can AI models code autonomous drones?” source ↗
• Robotics“Can AI models code autonomous drones?” source ↗

What it sells

• Evals“our evals Vending-Bench, Blueprint-Bench and Drone-Bench allow us to compare models directly in simulation.” source ↗
• Agent safety and verification“We study and deploy frontier AI in the real world to ensure organizations run autonomously by AI are safe.” source ↗

See it in the vendor universe →

Andon Labs on RL Research → Website ↗

Ooak Data sells RL environments built from anonymized real company data (digital twins of enterprise tools like Slack, Gmail, Jira) for training and evaluating AI agents on multi-step workplace workflows.

Why it's in Enterprise work

• Enterprise work leads with“We source real enterprise data, anonymize it into digital twins, and generate reinforcement learning environments with expert-level tasks.” source ↗

What it sells

• RL environments“We turn real company data into RL environments where AI agents learn to work in the real world.” source ↗
• Evals“Digital twins that let you evaluate agent performance against realistic company environments before going to production.” source ↗

See it in the vendor universe →

Ooak Data on RL Research → Website ↗

Deeptune builds RL training gyms that simulate popular software like Slack and Salesforce for frontier AI labs to train agents, and is now part of Mercor.

Now part of Mercor.

Why it's in Enterprise work

• Enterprise work leads with“We've built 100s of gyms that mimic popular software like Slack and Salesforce.” source ↗
• Coding“A simulation environment where AI can practice doing work (like writing code or making a spreadsheet)” source ↗

What it sells

• RL environments“We build training gyms for frontier research labs.” source ↗

See it in the vendor universe →

Deeptune on RL Research → Website ↗

Trajectory Labs provides frontier AI labs with expert red teaming, safety evaluations and benchmarks, and RL environments and safety data for training agents.

Why it's in Security

• Security leads with“we were the only external team to complete one of Anthropic's cyber tasks against Opus 5's safeguards.” source ↗
• Coding“How robust are coding agents to attacks they've never seen?” source ↗

What it sells

• Agent safety and verification“Our expert red teamers continue to find new vulnerabilities in every frontier model we've tested and are relied on by frontier labs to keep their models safe.” source ↗
• Evals“We built the Search-PI Complex prompt-injection benchmark in partnership with Meta.” source ↗
• RL environments“Create tasks, environments, and adversarial examples that expose consequential agent failures.” source ↗
• Human data“We connect the world's top red teamers, researchers, and engineers to mission-critical projects at frontier AI labs.” source ↗

See it in the vendor universe →

Trajectory Labs on RL Research → Website ↗

ARIMLABS builds production-fidelity RL environments and benchmarks for long-horizon AI agents in cybersecurity, SRE, compilation and STEM, plus LLM safety benchmarking.

Why it's in Security

• Security leads with“Cyber — offense and defense environments” source ↗
• Coding“Compilation — build, toolchain, and dependency repair” source ↗
• Science“STEM — verifiable reasoning and problem solving” source ↗
• Enterprise work“SRE — incident response simulations” source ↗

What it sells

• RL environments“Carefully calibrated training environments for the agents of tomorrow.” source ↗
• Evals“the benchmarks that measure what frontier models actually do.” source ↗
• Agent safety and verification“Testing for containment failure, deception, and hidden capability in frontier models.” source ↗

See it in the vendor universe →

ARIMLABS on RL Research → Website ↗

Gray Swan AI sells AI security products: Shade automated red teaming of models and agents, Cygnal runtime enforcement against adversarial inputs, and a crowdsourced red-teaming Arena, and serves as a pre-release evaluator for frontier labs.

Why it's in Security

• Security leads with“Shade runs attacks that work against your models, agents, and guardrails — in your deployment context.” source ↗

What it sells

• Agent safety and verification“It classifies and blocks adversarial inputs and unsafe agent outputs in production, with the highest block rate, lowest false positive rate, and lowest latency on the market.” source ↗
• Evals“As the leading pre-release evaluator of frontier model alignment, we deliver the best available active behavior assessment and enforcement for Enterprise AI.” source ↗
• Human data“Gray Swan tests against attacks discovered by 15,000+ adversarial researchers.” source ↗

See it in the vendor universe →

Gray Swan AI on RL Research → Website ↗

Incalmo builds RL environments that train AI models on realistic, verifiable cybersecurity challenges.

Why it's in Security

• Security leads with“RL environments that task models with solving realistic security challenges” source ↗

What it sells

• RL environments“One form of this infrastructure is RL environments that task models with solving realistic security challenges.” source ↗

See it in the vendor universe →

Incalmo on RL Research → Website ↗

Cua sells computer-use agent infrastructure (open-source desktop driver, cross-OS sandboxes and elastic fleets) plus a catalog of 4K+ computer-use RL environments, the Cua-Bench benchmark, and verified trajectory datasets.

Why it's in Computer use

• Computer use leads with“Train and evaluate agents on real computer-use tasks.” source ↗
• Enterprise work“CRM, ERP, collaboration, and self-hosted business workflows.” source ↗
• Coding“Coding, debugging, database exploration, and API testing.” source ↗
• Science“Scientific computing, GIS, statistics, and research software.” source ↗
• Media“Image, 3D, audio, video, and vector-editing tasks.” source ↗
• Games“Editor workflows, build debugging, and in-engine QA testing.” source ↗
• Chip design“The best frontier agent clears just 6 of 25 expert KiCad tasks.” source ↗

What it sells

• Sandboxes and compute“Elastic infrastructure that scales sandboxes into warm pools of machines you claim on demand for evals, RL loops, data generation, and batch rollouts.” source ↗
• RL environments“Cua has 4K+ environments and 400K+ tasks spanning desktop apps, cross-app workflows, browser surfaces, and generated state variations.” source ↗
• Evals“Eval and gym authoring for cross-surface agent benchmarks.” source ↗
• Datasets“We run the rollouts on the same environments and deliver verified trajectory datasets, packaged for your data-ingestion pipeline” source ↗

See it in the vendor universe →

Cua on RL Research → Website ↗

Chakra Labs sells deterministic, high-fidelity clones of production software (Notion, Gmail, Salesforce, Slack, GitHub and others) with GUI and MCP surfaces as RL environments, plus trajectory datasets and benchmarks for training and evaluating computer-use agents.

Why it's in Computer use

• Computer use leads with“Collaborative reinforcement learning environments for computer use research.” source ↗
• Enterprise work“Notion Figma Canva Gmail LinkedIn Salesforce Amazon Slack GitHub Grafana Google Calendar” source ↗

What it sells

• RL environments“Deterministic, pixel-perfect reinforcement learning environments with frame-accurate state control” source ↗
• Datasets“Deterministic environments with frame-accurate state control, high-fidelity trajectory datasets, and mixed-modality training capability.” source ↗
• Evals“Model performance on dojo-bench-mini leveraging Chakra environments” source ↗

See it in the vendor universe →

Chakra Labs on RL Research → Website ↗

Markov sells computer-use training data and environments, including screen recordings with synced mouse/keyboard events, CAD task environments with rubrics, and gaming datasets.

Why it's in Computer use

• Computer use leads with“Data + environments for training computer-use AI.” source ↗
• Coding“Browser useCodingDesignSpreadsheetsProductive workflows” source ↗
• Games“Screen recordings with synchronised mouse and keyboard inputs across Valorant, Minecraft, GTA, and more.” source ↗
• Enterprise work“Real workflows across Salesforce, Blender, Photoshop, and more.” source ↗
• Media“Real workflows across Salesforce, Blender, Photoshop, and more.” source ↗

What it sells

• Datasets“Rich computer-use data built around complete workflows, synchronized interaction signals, and verifiable outcomes.” source ↗
• RL environments“End-to-end CAD environments with the full prompt, reference inputs, complete screen recording, deliverable previews, and step-level action trajectories.” source ↗

See it in the vendor universe →

Markov on RL Research → Website ↗

EdotEnv builds multi-step RL environments from financial market data for frontier labs and academic groups to evaluate and post-train research agents.

Why it's in Finance

• Finance leads with“we build RL environments at scale from financial market data that increase in difficulty” source ↗

What it sells

• RL environments“we build RL environments at scale from financial market data that increase in difficulty, so agents can continuously hillclimb.” source ↗
• Evals“We work with frontier AI labs and academic groups building research harnesses, evaluation benchmarks, and post-training environments.” source ↗

See it in the vendor universe →

EdotEnv on RL Research → Website ↗

Halluminate builds benchmarks and RL environments for AI agents doing knowledge work, starting with financial services and consulting workflows.

Why it's in Finance

• Finance leads with“Our first area of focus is economically valuable workflows in financial services and consulting” source ↗
• Computer use“runs inside dynamic desktop environments, and spans trajectories that reach hundreds of steps.” source ↗
• Enterprise work“building benchmarks and RL environments to push the frontier of AI in knowledge work.” source ↗

What it sells

• RL environments“Halluminate is a data research lab building benchmarks and RL environments to push the frontier of AI in knowledge work.” source ↗
• Evals“We present Westworld Finance Diligence Bench, a benchmark of 88 problems that evaluates AI agents on a complete company acquisition due-diligence process.” source ↗

See it in the vendor universe →

Halluminate on RL Research → Website ↗

Dissei publishes financial reasoning evaluations and a leaderboard measuring how AI models reason from financial evidence, plus a public dataset.

Why it's in Finance

• Finance leads with“Studies in what models learn from finance, and where their judgment breaks.” source ↗

What it sells

• Evals“How AI models and LLMs reason from financial evidence, compared across seven reasoning categories and scored on a 0–100 continuous reward scale” source ↗

See it in the vendor universe →

Dissei on RL Research → Website ↗

Verita AI supplies frontier labs with a curated community of PhD-level experts who provide personalized, preference-based human feedback and evaluation data, especially for aesthetics and taste.

Why it's in Media

• Media leads with“Verita is building at the intersection of aesthetic dimensionality, individuality, and preference” source ↗

What it sells

• Human data“Our platform solves this by collecting open-ended information about every expert in our community and building on each profile as they work with us.” source ↗
• Evals expanding“we share what we learn through published benchmarks and research.” source ↗

See it in the vendor universe →

Verita AI on RL Research → Website ↗

Exabite sells off-the-shelf and custom-scraped pre-training, post-training and RL data catalogs to frontier AI labs.

Why it's in Media

• Media leads with“including video, documents, and global web text beyond Common Crawl” source ↗

What it sells

• Datasets“Move faster with a deep off-the-shelf catalog spanning pre-training, post-training, and RL—including video, documents, and global web text beyond Common Crawl.” source ↗

See it in the vendor universe →

Exabite on RL Research → Website ↗

AIChamp sells a vendor-neutral factory loss assessment and non-invasive digital twins of manufacturing stations in which robot-AI teams train and are tested against a plant-set target before a floor pilot.

Why it's in Robotics

• Robotics leads with“We build the twin, then robot-AI teams, PLC changes, vision and equipment options are tested in it.” source ↗

What it sells

• RL environments“Vetted robot-AI teams train on your task in the twin” source ↗
• Evals“Every option tested in a twin of your station before capex” source ↗
• Enterprise AI services“The fix is installed by your integrator, or by us or a local partner, then measured with the same sensors and method.” source ↗

See it in the vendor universe →

AIChamp on RL Research → Website ↗

Uber AI Solutions sells human data services (data collection, annotation, testing and evaluation) from a global network of over 10 million contributors and experts for generative, agentic and physical AI.

Why it's in Robotics

• Robotics leads with“Build robotic and autonomous systems to operate in the real world.” source ↗

What it sells

• Human data“Tap into our global network of over 10 million contributors and specialized experts spanning a range of industries, languages, and disciplines” source ↗
• Evals“accessing the best of Uber’s data collection, annotation, testing, human-in-the-loop validation, and evaluation capabilities.” source ↗

See it in the vendor universe →

Uber AI on RL Research → Website ↗

Hillclimb builds and sells RL environments and aggregated human research data to frontier labs to improve models' AI research capabilities.

Why it's in Science

• Science leads with“If you'd like your models to be better at AI research, email us at” source ↗
• Math“In our previous venture, we deployed a similar approach for math, fostering a dedicated community of IMO medalists” source ↗

What it sells

• RL environments“To further scale these ambitions, we sell the environments we make to frontier labs.” source ↗
• Human data“fostering a dedicated community of IMO medalists to co-train a state-of-the-art model” source ↗

See it in the vendor universe →

Hillclimb on RL Research → Website ↗

d_model is an AI research lab that builds RL environments for open-ended interpretability and alignment research tasks, partnering with frontier labs to train agents as researchers.

Why it's in Science

• Science leads with“partnering with frontier labs to turn their models into capable interpretability and alignment researchers” source ↗

What it sells

• RL environments“We build RL environments for open-ended interpretability tasks. These environments teach agents how to do cutting-edge research, not how to reward-hack.” source ↗

See it in the vendor universe →

dmodel on RL Research → Website ↗

Normal is an applied research lab selling data and evals to advance frontier AI capabilities in hardware design, covering rockets, robots, aircraft, satellites, medical devices and semiconductors.

Why it's in Chip design

• Chip design leads with“Normal is an applied research lab for hardware design.” source ↗
• Robotics“rockets, robots, aircraft, satellites, medical devices, and semiconductorswith data and evals.” source ↗

What it sells

• Evals“We advance frontier AI capabilities for thosebuildingrocketsrockets, robots, aircraft, satellites, medical devices, and semiconductorswith data and evals.” source ↗
• Datasets“We advance frontier AI capabilities for thosebuildingrocketsrockets, robots, aircraft, satellites, medical devices, and semiconductorswith data and evals.” source ↗

See it in the vendor universe →

Normal on RL Research → Website ↗

General Reasoning is an AI research lab that runs OpenReward, a platform for discovering and serving community RL environments, publishes long-horizon benchmarks like KellyBench, and sells BackSearch, a point-in-time search API for forecasting, evals and training.

Why it's in Games

• Games leads with“Every frontier AI model lost money when tasked with betting on Premier League matches over a full season, a new benchmark reveals.” source ↗

What it sells

• RL environments“An open resource for discovering and experimenting with community RL environments with simple API endpoints and autoscaled infrastructure.” source ↗
• Evals“A benchmark for long-horizon sequential decision making.” source ↗
• Datasets expanding“A frozen archive of news, SEC filings, arXiv, Wikipedia and the wider web, with point-in-time search and fetch for forecasting and reproducible evals and training.” source ↗

See it in the vendor universe →

General Reasoning on RL Research → Website ↗

Latch builds verifiable benchmarks that evaluate AI agents on biology and drug discovery tasks, such as single-cell, spatial biology, therapeutics and biosecurity analysis.

Why it's in Health and bio

• Health and bio leads with“Researching and ImprovingAI Agents for Biology” source ↗
• Security“BioSecBench-Refusal: A Paired Metric for Performance and Alignment in Agentic Biosecurity Risk Assessment” source ↗

What it sells

• Evals“We introduce TxBench-Oligonucleotide Discovery, a verifiable benchmark of 113 evaluations that tests whether AI agents” source ↗

See it in the vendor universe →

Latch on RL Research → Website ↗

Ulam sells research-grade mathematical reasoning datasets, private benchmarks and verifier-backed RL gyms to frontier AI labs.

Why it's in Math

• Math leads with“Original, unpublished private mathematics problems for training and evaluation—from AIME and olympiad level through graduate, PhD, and research-level mathematics.” source ↗
• Coding“train and evaluate agents in verifier-backed RL gyms for code, algorithms, science, and mathematics.” source ↗
• Science“train and evaluate agents in verifier-backed RL gyms for code, algorithms, science, and mathematics.” source ↗

What it sells

• Datasets“Use verified research trajectories, proof attempts, critiques, repairs, and failure traces off the shelf” source ↗
• Evals“Custom benchmark suites and eval protocols for labs that need private, difficult-to-overfit measures of frontier reasoning.” source ↗
• RL environments“Train and evaluate agents in stateful, verifier-backed research environments where progress, revision, and failure become learning signals.” source ↗
• Human data“Iterative traces where models explore, revise, and refine ideas with expert intervention at the points that matter.” source ↗

See it in the vendor universe →

Ulam on RL Research → Website ↗

Andromede is an RL data lab that programmatically generates RL environments, tasks and verifiers from real-world data for post-training and evaluating long-horizon frontier agents.

What it sells

• RL environments“Programmatic generation of RL environments for post-training and evaluation.” source ↗
• Evals“providing a scalable, repeatable post-training and evaluation pipeline for long-horizon tasks.” source ↗

See it in the vendor universe →

Andromede on RL Research → Website ↗

Modal sells a Python-native serverless cloud for AI, offering on-demand GPU compute, multi-node clusters, inference serving, and fast-booting sandboxes for agents.

What it sells

• Sandboxes and compute“VM Sandboxes are built for those who need to give their agents the power of a full computer.” source ↗
• Training platforms“Cloud infrastructure for teams that develop, train, and serve AI applications at scale.” source ↗

See it in the vendor universe →

Modal on RL Research → Website ↗

Daytona sells secure, elastic sandbox infrastructure where AI agents run AI-generated code in isolated, stateful environments for coding agents, computer use, evals and RL training.

What it sells

• Sandboxes and compute“Secure and Elastic Infrastructure for Running Your AI-Generated Code.” source ↗

See it in the vendor universe →

Daytona on RL Research → Website ↗

E2B sells isolated Firecracker microVM sandboxes for AI agents to run code, use browsers and desktops, and execute RL rollouts, hosted in its cloud or the customer's.

What it sells

• Sandboxes and compute“Coding, computer use, background jobs, RL rollouts: one API, an isolated microVM per session, in our cloud or your AWS, GCP, or Azure account.” source ↗

See it in the vendor universe →

E2B on RL Research → Website ↗

Metaphi AI builds training environments for AI models aimed at closing the gap between real-world conditions and the environments models are trained in.

What it sells

• RL environments“There is a large and widening gap between the real world and the environments models are trained in.” source ↗

See it in the vendor universe →

Metaphi on RL Research → Website ↗

Vetto AI designs learning experiences, feedback systems and evaluations for frontier AI labs to improve model capability.

What it sells

• Evals“Vetto designs learning experiences, feedback systems, and evaluations that directly improve real-world model capability.” source ↗

See it in the vendor universe →

Vetto AI on RL Research → Website ↗