Back to list
Industry NewsAsanaOpenAIAI Agents

Asana Cuts Browser Agent Costs 76x and Boosts Speed 5x in Tests Powered by OpenAI Codex

In recent browser agent testing reported by OpenAI, Asana achieved a 76x reduction in model costs alongside a 5x increase in operational speed. By integrating advanced model capabilities through Codex—identified in reports with GPT-6 Astra and GPT-6.1 Sol—Asana tested enhancements to its browser automation workflows. These performance gains demonstrate substantial improvements in efficiency and execution speed during internal evaluations. According to the announcement, the primary objective of these architectural tests is to allow Asana to offer more capable AI models to its end customers without incurring prohibitive operational expenses. The published test results highlight how optimizations in model integration can drastically reduce deployment barriers for interactive agentic tools in enterprise environments.

OpenAI Blog

Key Takeaways

  • Significant Cost Reduction: Asana recorded a 76x drop in model costs during evaluations of its browser agent.
  • Substantial Speed Gains: The tested browser agent executed tasks 5x faster compared to previous implementations.
  • Codex Integration: The test benchmarks utilized model integrations within OpenAI's Codex ecosystem (referencing GPT-6 Astra and GPT-6.1 Sol).
  • Customer-Focused Scaling: The efficiency gains are targeted at enabling Asana to deliver more capable AI models directly to customers.

In-Depth Analysis

Drastic Cost Efficiency in Browser Agent Testing

According to findings published by OpenAI, Asana's browser agent tests demonstrated an unprecedented 76-fold reduction in operational model costs. Browser agents inherently require continuous contextual evaluations, DOM element interactions, and sequential decision-making loops, which typically consume substantial model compute. By utilizing new model implementations in Codex, Asana successfully compressed these computational expenses by 76x, marking a pivotal milestone in making agent-driven automation economically viable for continuous evaluation.

Acceleration of Agent Execution Speed

In addition to drastic expenditure reductions, Asana recorded a 5x improvement in response and execution latency during testing. In web-based agent environments, latency directly dictates the viability of automated workflows; agents must parse page structures and trigger events within actionable time windows. Achieving a fivefold increase in test execution speeds addresses latency bottlenecks, allowing browser agents to carry out multi-step tasks much faster.

Delivering More Capable Models to End Users

As outlined in the source documentation, the driving motivation behind these tests is to provide Asana customers with access to significantly more capable models. Traditionally, deploying more intelligent and complex agentic models introduces severe trade-offs in operational cost and runtime latency. By demonstrating both a 76x decrease in cost and a 5x increase in operational speed, the reported tests indicate that higher-tier intelligence can be introduced to end users without the standard performance or financial penalties.

Industry Impact

Advancing Practical Enterprise Agent Adoption

The documented efficiency metrics highlight practical advancements for agent deployment across enterprise productivity software. High inference costs and slow execution times have historically formed the two biggest impediments to deploying autonomous browser agents at scale. Achieving a 76x reduction in cost alongside a 5x acceleration directly addresses these hurdles, illustrating a viable path for productivity platforms to embed interactive agents into production workflows.

Setting New Benchmarks for Model Integration

The collaboration between Asana and OpenAI's Codex architecture underscores how tightly integrated tooling and updated model generations can optimize agentic workloads. As reported, using these specialized model environments enables organizations to sustain intensive automated browser tasks at a fraction of standard operating expenditure, setting a concrete technical precedent for future agent testing and deployment across the AI industry.

Frequently Asked Questions

What efficiency gains did Asana report in its browser agent tests?

Asana reported that its browser agent achieved a 76x reduction in model costs and operated 5x faster during testing.

Which models and platforms were used in the reported tests?

The tests utilized model integrations within Codex, specifically referencing GPT-6 Astra and GPT-6.1 Sol as detailed in the release.

What is the goal of Asana's performance optimization tests?

The stated objective of the testing is to enable Asana to offer more capable models to its customers while maintaining operational viability.

Related News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini
Industry News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini

A Google Research Blog entry titled “How Google Duplex evolved to also call businesses for you in Gemini” describes its subject as the evolution of Google Duplex to support calls to businesses on a user’s behalf in Gemini. The entry is dated November 20, 2026, and the supplied article text consists only of the label “Generative AI.” The headline identifies the capability and connects it to Duplex, but the available text does not explain how the system evolved, how users request a call, or where the feature is available. The headline establishes the topic, but the supplied material does not provide a technical account or rollout details.

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis
Industry News

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis

TechCrunch AI reports that AI claimed a solution to the Riemann hypothesis, a mathematical problem the headline describes as 167 years old. A human also put forward a competing solution. The report, written by Tim Fernholz and published on November 20, 2026, describes mathematicians as astonished by the AI-driven result. The supplied excerpt does not explain either proposed solution, identify the AI system or the human claimant, or say whether either claim has been verified. The distinction between a claimed solution and a confirmed one is central to reading the report: the excerpt establishes that two claims are being reported, but it does not establish that either has resolved the hypothesis.

METR’s AI Time Horizon Update Highlights Exponential Gains and a Gap Between Judgment and Coding
Industry News

METR’s AI Time Horizon Update Highlights Exponential Gains and a Gap Between Judgment and Coding

A new report from METR, an AI evaluation nonprofit, updates its time horizon plot and reports two findings about model capabilities. Some AI capabilities continue to improve along an exponential trajectory, but models do not perform as well on tasks that require judgment as they do on coding tasks. Grace Huckins covers the update for MIT Technology Review. The two findings have different scopes: one describes how some capabilities are changing over time, while the other compares model performance across two kinds of tasks. The available excerpt ends before it explains the plot in detail, so it does not establish how the results were measured, how large the performance gap is, or which specific capabilities are following the exponential trajectory.